EPISODES
Episode 27.
Stefan Tsvetkov , Founder at RealtyQuant
All right, all right, I’ll be honest. I think I’m a pretty intelligent, well rounded human but I have to op the Auntie for this week’s guest, Stefan’s fat cove, investor and founder of Realty Quandt he is an incredibly intelligent human. I feel like we in the investing world focus so much on the up side, but what about downside analysis? Join us as we dive into this and a slew of other metrics on data driven investing. Thanks for listening. You’re listening to the real estate of things podcast. Welcome to the real estate of things podcast. I’m your host, Dalton Elliott. Step on Spet Cov, founder of realty quant. Thank you so much for joining me today. Thank you for having me, though. It’s an appreciate it absolutely so. You’re based in New York, your financial engineer, you’re a multi family investor, you are a busy man, my friend. So I think let’s kick it off. Just tell us a bit about yourself, tell us about realty quant, and then I want to dive into the data side, the real estate side, but tell us a bit about your background and the company. Yes, absolutely, and so I’m used to to APEON. I came to the State of twenty two a king here for my musters. So studied financial engineering in New York City and I used to work in finance for about for about a decade and I transition to being a fool time investor in the the past couple of years to have been doing that. and My company, reality quantity, is really data analytics, education and kind of technology platform. Kin, I got it. And how long is realty quant been up and running? Reality quants being I would say I’ve been focusing on that the most in the past one year, even even less. I’ve had like bits and pieces. Reality Coal is really the product of kind of my own investment, own investment kind of effort. So I would say I’ve had like many of the pieces to it for probably like as many as two three years back, but I’ve kept kept most of them in house. So it would write like my own like scripts and, let’s ee, like python scripts for some of your alady and if the audience, if they’re acquainted, and I would just write like all these codes and how like thousands of lines of code to find the use, to find markets, you know, things like that. Then kind of like the model started building on top of each other and then I’ve started more recently, in the past one year, releasing some of them as products out for other investors to use. Got It? That so you have you have something special there. You have a special, special viewpoint, and you know, before we have to own. I was thinking about kind of reality, Quan, and the fact that you are an active real estate investor, and thinking about that, thulins, of what I do on a daily basis. Our company does line capital on the day job. And let me ask this. Would you say that your investment philosophy is void of emotion? Is it just the numbers have to pan out in the algorithm? And you know if that can be a yes or no question? Maybe I can’t. Maybe I can’t, but then really talk to me, give me a peek under the hood of kind of what data points you’re looking at, how you’re scraping those data points, how’s everything work? Yeah, this is a great question. So if it’s a yes or no, I would say no. So I should definitely not devoid of emotion and kind of I do review like Odius afterwards, manual and so forth as well. But just like to give a background. What is that’s a driven real estate investing. It’s basically, let’s say, for a person like me coming out of finance, it’s basically trading the private true estate market kind of like stocks and trying to proxy and get a schools as pope possible to if it was a public market,…
…which it’s not, and they’re like what’s of advantages to that? Obviously and like to capture like all that inefficiency that is there, but trying to build your own investment systems to to a level that it’s close is possible your trade like stocks, and what that really means is like several tanks. So I can give example of what to me is that to domn investing. It’s if we take on market in a most investors and interesty and off market listing things and use of course, but if we take, just for the completeness, on market that will kind of under right own market data in either commercial or residential, end to end. So you have like the full and the writing, have a recall your cap rates, cash and cash returns for your financing terms and so forth, and so then you have some off market modeling and collecting of data. So let’s I would scrape like multiple lack of market commercial deuse of facebook and other social media and then kind of those are harder to automatedly underwriting an automated way, but kind of has have these data and then combine it with maybe can acquisitions analyst who would would underwrite those. So that’s kind of has is another thing. So off market commercial modeling in sectually get gets very interesting and there is a methodologytube for modeling Commercial Moulti family, which I developed at my firm, which is utilizes rental listings date and sort of does a preliminary income expense sheet analysis of a property and can do it. It’s scale for like any market in the country and so forth. It’s a very good too, and that’s that’s one thing I’ve been using as well. And then another component of these u investings are and cap statistical regression analysis for markets to arrive like the best appreciation can of running like appreciation forecasts in a very rigorous way rather than just, you know, picking like fundamentals that drive values supposedly, but doing it like a rig rigorous regression, and it’s as part of that. Measuring downside risk is another one. So down cent risk. That’s like very I feel like in the realest in industry. It’s not commonly used and it’s also the industry wherey thing is the easiest one to use. So it’s easiest, one of the easiest things, to measure downser risk, because it’s a very fundamental asset. It’s driven by population, income, housing supply, so one can measure that. So this is another piece. Automated Valuation Models, I would say a fort one. So things like zero’s estimate and so forth, with all the caveats of that and limitations and so forth. But in highly liquid markets that has helped me discover the use. Like places like upstate New York and you can run your own avm and try to find like properties that are like very, very mispriced, and it happens to be. Sometimes it’s rare, but that’s that helps you to detect it. And then using like some natural language processing for textual quasification for various things. So one would be conditioned property conditions, so sort of beat on market roof market properties, doing textual quasification, which is the really a machine warning method. There are various machine warning methods to actually quassify and be able to process through your listings and automate automated under right them in a more efficient way. And then another thing would be what something about the company in New York Code Foxy Ai do like and I don’t do it myself. It’s image, image classification and condition scoring for images. This is another important component for data driven investing because if you are able to actually go through a real estate images in an automated man or well that can save it time, timing. You don’t need like, you know, having like a nonalys to do that. Then they can look at like the already under a written deals and sort of do deeper, deeper dive into them. But but again, like those kind of a long winded answer, like going through like so only different topics. But really like to your question. Is it removing emotion of investing? No, it doesn’t. In fact, I’ve been working right now and hiring…
…an acquisition snalyst because the moneyo component is still there. But it just helps you to scale a little more and have more insights and have more preliminary underwriting than you would otherwise have. Now that makes complete sense. There’s you know, he just technology isn’t to the point where you can fully remove the human touch on on most things, so that makes sense. One thing you talked about was at the downside analysis. It’s so what’s so? Talk to me more about that. I think you know the masses. You know everybody. If you look at stock market since inception, if you look at American real estate value over the last a hundred years, like up into the right, but we know that there have been some significant and elongated down turns. So I’m curious about that because, yeah, we’re always looking kind of forward flow. What are the metrics for growth in this market? You know, is Amazon moving in? What’s going on? But it’s not often that you hear kind of the more risk side approach on it. So what goes into projecting potential downside? So you agree? I agree, and this is actually this a great question. Yeah, we often hear about all the appreciation predictors, right, and there isn’t like there’s like a risk. Manage one side so much, I feel, and then so, so, yeah, so what I what I myself did. So it so the reality quanted the beginning of Covid so I always like super concerned. If you know, like what if real estate takes a down turn? You know, are Wat to peak? You know the kind of the usual concern that many people have, and and so I looked up like different studies and like. So one thing that inspired me, it comes from finance. It’s a study code measuring the bubble by Hausman Investment Trust. So this is one thing that I worked up and sort of has like different ratios and different metrics to try to see, like how much you know the stock markets is overworld. And admittedly it’s like way harder to arguably so it’s way harder to do it in the stock market because, you know, it’s maybe some of your audience knows, especially in recent years, the stock damp five hundred, or the main index in the stock mark is very driven by the top five technology companies and some other technology companies and technology companies there. It’s very hard to have a valuation there, right. So you don’t you don’t know what they should be word. So so I don’t have, for example, this analysis for the stock market, and that’s actually one reason why don’t own any stocks now, because if you like, if you don’t know valuation, you’re actually not a professional investor. You’re not like to the level of what what I’m were on Buffett was in the s or something like that because you you’re just constantly guessing where the market is and and they can be guessing if the stock market is overvolowed or not, but they don’t actually know and they don’t have exact measurements and that, and so the same effort in reverse it. So that’s like one thing that inspired me. Husman Investment Trust, John Husman, Ph d in finance, a thing. It’s like who heads that Hedge Fund? He has some studies. Then I woked up, for some of your audience who know, Neil Bauer. He uses a vendor cold wocal market monitor for some market data and that’s run by INGLE WINDSOR ING winsor was in two thousand and five and two thousand and six on CNN and was sharing the certain markets in California and California and and for the primarily he was focusing where dangerously overvalrowed, and that was kind of two years ahead of the actually the downturn that happened in two thousand and seven and so and he was doing. What he was doing is like just simple pricing, come ratios and kind of derived from that. And so that was able actually to predict, you know, some of the overvoluation that time quite well and and yeah, so it. So I went on to do similor similar study on my end. So I just my predicted variable was the declines in different states and different counties after the global financial crisis, and so I went on like what predicted that the best? They looked at for closure rates, it looked at like various different metrics. I looked at the price utility, like Chriska, just the returns in different regions and things like that,…
…and they weren’t very predictive. And what actually like seem to work the very best is well, it’s really real estate prices are driven by income, population and housing supply broadly speaking, and so if you take a predicted price of those three, like we are regression or whichever, met at a predicted price of those three fundamentals and then if you look at deviation versus at predicted price on some historical window, that is actually working, which is a if you think about it’s a really simple to even one side and that works really well. And so so I started first with I just took pricing come ratios in all different regions. They took them over a moving average twenty year time window, and then I took like when has been the peak in each different market, like county, state, can be, on a zip code, etc. You know, the peak of the prices and where was this deviation from? Essentially, we can say pricing. Come racials. It’s affordability, right, so it’s like what home prices are times the personal income in our region. And so this kind of deviations in affordability, where they were in the different regions at the peak of the wob of financial crisis, and then how that correlated to the subsequent declines in the same regions, right, and so and so this study showed and like eighty five percent correlation at the state level to the subsequent actual decline. So it was very strong. So I could seem like, obviously this is not something I have invented. I have seen such studies like done by boom work economics and so forth, but they never knew like a sense of actually the actual predictive power of it. And that, if you was my ball at the time. And then I tested at the county level and I was like about seventy five percent, kind of way worse, you know, not so strong correlation in the county level, but still still high. And so and so I took that and like I thought, okay, that’s making comfortable. It did we like different methodologies, like, like I mentioned, incorporating population and housing supply and for acial based regression, based like several different ways, and then I kind of like I started tracking this and like computing like a where is it now, you know? So started at the beginning of COVID. Then the interesting thing is at the beginning of covid actually US real estate was fairly valid, and that was some different, if you will, like you know, cademics or like analyst, who who track that? There aren’t too many, but there is. There is a hitting, a studied for the Atlantic University, and then there is boom work economics, which I mentioned four countries, and so boomerk economics was saying like since two thousand and nineteen US estate is fairly vowed, and then they released it in the first quarter of two thousand and twenty one and it was still fairly voted broadly speaking. But then, and that was kind of an aligned with what I was seeing as well, it’s like around a hundred percent. Actually, it’s quite interesting when, though so many people for many years ay it’s over valid, but it was not at the time. And so and so I went to like and in actually in the there was an exception. So the big exception was the state of Idaho, which was like twenty five percent, twenty five percent over votewed. And boy seeing the boy see if some of your audience may have heard like this markets. I Cole Voice is the best performing city in America in price appreciation. So it really appreciates, really booming, you know, it does really well, but it’s also overwaowed. It’s so it happens. So boys who was like a thirty three percent, and that’s already like around like March two thousand and twenty or so, so forth, like around the time, like the end of first quarter of two thousand and twenty, and and yeah, and so it kept tracking them like quarterly, and so it’s can be stay the same. So let’s say, like I’m working and let’s see Forida and Texas, and I have this study for like Twozero, seven hundred US counties, you know, and so I tracked them like every quarter. And let’s say for it and Texas stay in like eight to ten percent, slightly overwalled range. Kind of fairly validays you know, they’re strong. I mean they’re various markets in there, obviously, but generally very strong states with many strong markets in them.
And and so they were like very consistent for four years, eight to ten percent, and a few Western markets like Arizona and Courado perhaps like thirteen to sixteen percent, something like that, kind of a little bit more over, worrowed and then, idea, who it? It’s twenty five percent, and and that, and and then in case. So that’s where it was. And at the second quarter and third quarter of two thousand and twenty one this dis changed and I came out at my web in or and they said, okay, it looks like this is the first indication of real estate bubble, because everybody was really happy about inflation. Kind of okay, we’re having in Flash. Inflation helps hard assets and that. But then again, that’s not to scare people off. It just like first indication in the sense, and what I mean by that just in the sense of okay, if we have four then Texas at day to ten percent consistently quarter after quarter for four years, will suddenly did doubled like seventeen eighteen percent, and either who which was a twenty five went to forty seven, and Arizona, which was a like Arizon, which was it like sixteen, went to like thirty, thirty one, and so kind of valuations kind of doubled. The reason for that is not complex at all. We have seen it first hand. It’s that asset inflation was very high, but some of the fundamentals inflation, like kingcome, did not catch up. An example that would be Arizona prices in the first half of last year group by seventeen percent, but income group by just one percent. So that’s kind of what drives this. Again, like this is not being barish on when those markets because they they actually have the highest appreciation forecast in my model as well, if we stay under the same cycle, and that is actually a very interesting concept, like perhaps a for your audience, because if a market is over varrowed, that’s not it’s not like you you’re predicting less appreciation for it. It’s quite the contrary actually. If it’s like so much over wart, it probably has very strong ter and very strong momentum that’s effected to continue by virtually any statistical model you can use out there. Bach Arima, facebook profit or I would say, or like exponential triples. Mooting those account like three methods one can used to forecast prices, let’s say. So that’s it. By whichever one one would use, it’s gonna gonna be. It’s gonna show significant growth in in Arizona. It’s going to show significant growth in either who, even if I the WHO is, you know, I would say over worrowed. But again, the moment that’s because we’re under the same market cycle and real estate is trend even so, how it tests if it’s trend driven? There’s something called like autocorrelation. autocorrelation tests, so you test if last year price growth correlates to this year price growth, and in most use regions it’s very high in all. Alaska was an exception, I can ask you, has like negative auto correlation. So there’s like no rhyme or reason there. It doesn’t make sense or it’s not like, let’s as if last year good performance predicts a not good one this year. It’s something generally it’s not the case. If we takes something like for them, autocorrelation was like seventy seven percent and and that so. So this kind of like to theirs trends. So their strength. So will things are booming and going well, like now with inflation. I’m not coming out saying, Oh, I’m negative on do aster money. They’re really, really well. They just one needs to have simultaneously be aware of the downside risk that once simultaneously carries with this same upside that you’re capturing. And so this, this is kind of the analysis there and then just becomes a question of do you want to go now? How do you pick markets? That’s it, based on this. Right. So you can go in the Western markets still or in invest but if you’re aware of the valuation, can kind of be tracking and try and okay, maybe taking a higher risk and trying to exit within two years or whichever, whichever timing thing is going to be right, because timing is uncertain obviously, and and and so that. So or you can go in under word markets if you’re super risk averse. And and actually what happened in during goal of financial crisis in under varied states is very interesting because there were ten, ten states at the time, including texts, which were under vowed, so negative valuation metrics in the same analysis. And so…
…they the other the drop that they had was four percent on average. So, for example, of Texas drop was also four percent, and that’s very interesting because the median income drop in the US at the time was four percent. So if he’s saying like kind of price normalized by income terms, and it’s not only income that drives it, but but let’s say, as an approximation, price normalized by income, turns actually didn’t drop. So it’s kind of like see valuation stay the same. It’s very interesting. So that’s kind of you have the big goal of financial crisis where you know, places like California, Arizona and Evald and for a drop forty to sixty percent, but nevertheless the under varrowed states, they they stay the same kind of invaleration terms, and that is very, very interesting. And so this is this is telling me, okay, that seems to work in a way. That okay. Why does it works a well, it’s because it’s rested so fundamental. I guess that’s my own explanation. And it’s kind of driven by those fundamentals. And and yeah, yeah, that’s super interesting. So until you clarified it towards the end, when I heard over value, I have my immediate reaction was like no, bad, not good, but it’s fine, as long as he said, the current market conditions and assumptions stay true, and then whenever a wrench gets thrown into one of those, then you have to go back to the drawing board. So it’s it seems to be a pretty good to dominate down for my simple brain. Right. Seems to be a good risk reward kind of measure. Right. Yeah, I agree. I agree. Yes, it’s a risk crew wor thing. It doesn’t predict your appreciation. I always feel like if you say something like over, it’s if like one would expect that something underwatered. It’s very it’s just depends relative to what it’s relative to fundamentals. But it doesn’t. That’s being under one of the relative to fundamentals interesting enough. It just doesn’t drive more growth because it can stay under about it for really long. Actually. It’s quite interesting and like many of the under one would markets will they they often are depressed, and so the I’m not predicting that they appreciate, but they neverthless do have very well downside risk. That’s something interesting because, like they’re many people I have heard, you know sometimes on Linkedin they may say something like Oh, by Houston assets or something like that. By this kind of like strong markets assets and you’re going to be okay. Well, that’s not how downside risk works, because downsay risk derives from kind of an irrational overallmotivation relative to from them, and that’s at least what they see. It doesn’t derive from simply better risk adjusted returns, like price relative to Vutility, even because I tested that. It doesn’t derive even from all these markets. Are More of water or less? What? Neither. Even that doesn’t correlate well. And so it really derives from like wall when there is kind of an irrasstional component and will, where is it most likely to happen? Often in strong markets. So that is actually the case. Now there there are some exceptions or some interesting cases there. So, for example, Denver or alum at the beginning of COVID was one of the most best performing cities. It was fairly volved, was like top five price performance and its valuation, in at least in my model, was at zero percent. Was Fairs bowed and it went a little bit of thin. It like twelve percent, kind of overwater then with inflation and so forth. But it doesn’t look doesn’t look that bad, and so it was actually fairly varved and have done really well, then there can be if he takes places like the State of Indiana, the State of Indiana was under varied and it kind of like those midrange markets that intuitively investors know that are not depressed. It’s not okay, it’s not like the northeast, you know, really depressed that nobody wants to invest in a way, but it is also, I mean except around some big cities, but broader. Is a broader region, but it is, you know, it had solid appreciation in Indiana over its state and certain markets it really well, but also was undervalued and now it is, let’s a fairly arts. That will kind of like an example of like what you said, the kind of like Chris…
…risk reward tradeoff and like deciding where do they want to go. So and so one can say, okay, want to go in this kind of medium up, medium, strong, medium, solid performance, really solid performance, but not like super strongest in the Western markets and southern markets, but on the other hand it’s under under too, fairly valued, and so you can kind of measure your your race in really this question is much more elemented towards the end of the market cycle and it wouldn’t so much matter, you know, at another time you would only be Workin at appreciation predictors and you would be aware of your evaluation metric, let’s say, what it is, but it’s not going to be so, so important and just starts to matter more and more towards the end of the psycho and and that. But yeah, it’s not an appreciation predictor. Your appreciation predictors are the same ones that, like every syndicator uses that every well, but I can talk to that. By the way. If you go on like the fundamentals and you choose let’s say population growth, job growth, like different things, and you try to kind of infor and you buy your picture market based on that, and if you try to infer appreciation, essentially your kind of inforign appreciation. Now if there’s all kinds of other factors, since you you like your properties or just like qualitatively the markets and so forth, but purely let’s say you’re trying to infer appreciation. And so when you do this, I did actually a study of that. So I took population growth, in can growth, housing supply change, and I try to predict those and then off those I predicted the prices. And then, on the other hand, they did another study where only it took the historical prices and just predicted the prices themselves, and so the second model had a five time smaller error. Actually, it actually focuses prices really well. If you take something like the state level, while we’re in the same trend, like it was two thousand and eighteen, two thousand and nineteen, the error in the price prediction was only one point four percent. So on is able to forecast prices to which is I’m not a fan of focusing price, to be honest. I’m much more like this kind of like downsiders can freask management side. But but again, because it’s because, I mean, when the trend deviates and changes, then so going to break, like your models being a break. But while we are in the same trend as had like one point four percent, are in two thousand and eighteen and two thousand and nineteen, rough free speaking. But if you go there with a cook and the syndicator approach or like investment man investor approach of actually doing the fundamentals, which is taken nice narrative to share with your investors and so forth, the error was seven percent. So way, way bigger. So like seven percent of the price is kind of big, and so it’s quite interesting. So actually that’s like something that I spoke like to some syndicators or like just investors that I know, and like I thought, I’m okay, it’s actually if you just go on looking at the historical price it because, seriously, is so trend driven, you would predict your short turned a one year head appreciation better. And so that’s kind of likes fires appreciation forecasting. But the downside, risk is very easy. Fundamentals. Deviation from fundamentals has like very, very strong correlation, very strong correlation. My my goal with this podcast is to have people on here who are smarter than me, and I’ve one on this episode, Steffan. Thank you, though, and that’s about that’s humble to so you get two points for that, my friend. So tell people if they want to learn more about realty quand where do they go? Tell us. Yeah, so, yeah, Rio ty Quandotcom all is the best way to reach to me. We have market data there and I’ve been working a few other products like commercial mounti family generation and some others. So rioty quandcom, that’s really the best way they can also find me on Linkedin. Beautiful, beautiful Stefan’s fat COB. I can’t thank you enough. I’ve thoroughly enjoyed this you. I have a scratch pad full of notes over here and I’m going to be probably coming back through this episode again to fill in the gaps of what I missed. This is really great education and a novel way of kind of viewing real estate. I’m a fan of…
…anything that we can take from kind of the emotions to the number side of it, and you know, certainly have to have that human element in there, but what you’re doing is it’s quite wonderful really on the real estate side, great approach. So thank you so much for joining. Thank you as well. Thanks everybody for listening and take care. Are you a real estate investor looking for the right lender that can finance all your deals and help you scale Lima one capital has the best suite of loan products in the industry, barnt, whether that’s fix and flips, fix and holds, building new construction or buying rental properties. They have incredible financing solutions for it all. A reliable common since Linder is one of the most important parts of your investment team, and that’s exactly what you get with Lima one. Let Lima one capital show you how they’ve helped thousands of real estate investor scale and increase their wealth. Check out Lima onecom or call eight hundred two five nine zero five ninety five to speak with the consultant and preparation for your next project. Thank you for joining us today on the real estate of things podcast. Subscribe and tune in weekly for new content from the industry’s best while we continue to unhack the nuances of this dynamic market. Follow US across social media for additional insights and analysis on the topics covered in each episode, and remember to rate, review and share the show.
The Real Estate of Things 