Housing Returns and the Emergence of the Safe Asset, 1465–2024

>> John: Okay, we're very happy to have Paul Schmelzing here to speak to us today as economic history is so important. I just love that you're doing this in a way that it's impossible for us to do. From Boston College and visiting the Hoover Institution. The name of your paper is Housing Returns in the Immerse of the Safe Asset.

 

This is, get this dates 1465 to 2024. You know, 1465. Anyway, go ahead, floor is yours.

>> Paul Schmelzing: Thank you so much, John and everybody for having me today. And the last time that I presented in this group was two years ago. And in retrospect, that has been one of the most helpful presentations.

 

And so I take that as a good omen. And today, I'm presenting a new project. This is the first time I'm really getting into this new asset class, if you will, housing. But in some ways, it's the logical next chapter from the previous presentation that was on real interest rates over space and time.

 

And so let's kick it off perhaps with this first chart here that many of you will be familiar with. It's simply Bob Shiller's real home price index that he keeps updating on his website. And that was one of the prominent charts in his book in 2015. But you can access the data and he keeps updating this on a monthly basis.

 

And it starts in 1890 and goes all the way to August 2024. And we see of course that if we start from the very right hand side, we see the 08 crisis pretty dramatically standing out here, but then we see this dramatic bounce back after the correction, right?

 

And in fact, if you go to his most recent updated data, he's basically saying that we're pretty much not only back at the 0607 peaks, no, we are way above these speaks in real terms, okay? So the index stands at 220 something. Which of course begs the question after all of these interventions, after all of these explanations about the 08 secular boom in housing.

 

If we are not seeing a continuation of that boom that started in whatever paper and book you take, most people would agree it started in the late 80s or somewhere in the mid-1990s perhaps. But it looks like in some ways a longer continuation of these real home price appreciation trends, right?

 

And so the interesting thing now, if we go from the right hand side more to the left, is that prior to 1970 or so, on the US basis, it doesn't look like much was happening, right? It looks pretty stagnant. And in fact that's explicitly how Bob Shiller characterizes the index between 1890 and say 1965 or 1970.

 

 

>> John RYLOE: Ken has a question, sorry.

>> Ken Judd: Say home prices.

>> Paul Schmelzing: Yeah, correct.

>> Ken Judd: How much of that is land? One reason I said, is because locally we know that Silicon Valley has extremely high home prices.

>> Paul Schmelzing: Correct.

>> Ken Judd: But that only happened in the late 70s when land became scarce.

 

Prior to that they just ripped up some orchards and put homes in. And then when land got scarce, that's when the housing price went up here. So I'm wondering how much of this is land?

>> Paul Schmelzing: That's a great point. And we'll talk a little bit about the decomposition as we as we go on.

 

But this is obviously for residential single family housing in the US, and so every part

>> Ken Judd: I mean the parts today are bigger.

>> Paul Schmelzing: Correct, and quality is a big impact. And we'll come to that. For the moment, I just want to emphasize the fact that this index, and we can look at the index for other advanced economies, they look pretty similar.

 

It looks pretty stagnant prior to that inflection point, whether you wanna pinpoint it to the next.

>> Speaker 4: And there are some big movements though. I mean, you're going from almost 130 down to 70, which over the course of 15 years, that's quite a drop. And then it recovers.

 

But it looks more stationary. But I would still characterize it as having quite a bit of volatility.

>> Paul Schmelzing: No, that's a great point. And we look at some of the volatility, including in a global context. So, this kind of long term secular evolution is pretty much consensual in recent literature.

 

So maybe one of the most comprehensive long-run house price projects, empirical projects, was the Knoll Schulerich part of the JST database that some of you will know as well. They reconstructed home prices in 17 advanced economies since the year 1870 and pretty much reached similar conclusions as Bob Shiller reached in his Long Run House Price Index.

 

Namely that something big happened in advanced economies in housing as an asset class around the year 1965 or 70 and prior to that, not much action, not much to see, essentially. Okay, there was one question for this and for your data.

>> Speaker 5: Will you look at listed prices, or transaction transacted prices?

 

 

>> Paul Schmelzing: So I try to replicate the Case-Shiller Index as closely as possible over centuries. And I'll talk about some of the deficiencies of the existing Case-Shiller Index in just a second. But it will be based on repeat sales of residential single house essentially. Okay, is there an attempt here to correct for the size of the house, the quality of the household.

 

To the extent possible and credible, will do, exactly. Okay, so the debate is very intense and very active beyond this stylized fact that a lot of the house price action started in the 1970s. Actually, a number of very important papers appeared over the last two, three years or so that took aim at long run house price trends and tried to improve on stylized facts.

 

So one of the big debates decides that long run trajectory concerns the performance of housing as an asset class relative to other risky asset classes, namely equities. And the same team that reconstructed long run house prices for these advanced economies claimed in the follow up paper that actually corrected for volatility.

 

Housing beats every other asset in the asset universe over the long run, including equities, okay? In real terms. That new stylized fact based on this new housing data has also caused a lot of pushback. And the debate is very active in the finance papers and beyond in more general interest papers.

 

But one big underlying methodological point that I will focus on today, is that Whatever of these new publications you use, whether you use the Case Shiller index and the series I just showed you for the US, we have major methodological issues with all the data prior to 1950 or so.

 

And it's very problematic to base this long run, secular, stylized facts on the current data that we have. So there's one major empirical point that I'm trying to convince you of today. Specifically for the US if you look into the sources that Bob Shiller actually uses for pre1950 data for the US, okay?

 

He uses a publication by Graeber, Blank and Winick that came out in 1956. Basically, it is a survey of current homeowners that asks current homeowners to estimate the value of their own home, okay? And then this survey data is spliced with the quality adjusted, much more reliable modern data after 1960 or so, okay?

 

To construct this long run.

>> Speaker 4: What's the nature of quality adjustment built into any housing types? Right, so the double counting, you have a separate. Does everyone with that?

>> Paul Schmelzing: The Case Shiller index is famously is a repeat sales index. So it takes the same identical house and-

 

 

>> Speaker 4: Well, that's an issue because there could be-

>> Paul Schmelzing: Correct, it tries to identify identical houses in terms of quality, okay? And then it basically says for each house where at least we have two transactions, we can construct this change in the price with basically constant quality, okay?

 

As long as these two points are relatively close together.

>> Speaker 4: You said something about quality adjustment though.

>> Paul Schmelzing: Yeah, so this is one way in the literature to address the quality change over longer periods of time by using the same house, okay?

>> Speaker 6: And that presumes an access, for example, of renovation.

 

 

>> Paul Schmelzing: Correct, there will still be quality changes, absolutely. But at least we can control for some of the big changes in housing as an individual unit that takes place over time, okay?

>> Speaker 6: Use our current example, in this neighborhood, land, as we pointed out, is very expensive.

>> Paul Schmelzing: Yeah.

 

 

>> Speaker 6: People immediately tear the house down and build a new one. So when we talk about renovation in terms of the value of the house, especially if you're trying to control the land value.

>> Paul Schmelzing: Correct.

>> Speaker 6: It could be quite significantly different.

>> Paul Schmelzing: Yeah.

>> Speaker 6: And if you're not controlling for that.

 

 

>> Paul Schmelzing: In such a case where the entire house is teared down and rebuilt from the ground up, for instance, that is a transaction that would not be included in the Shiller index that I would not include in some of the data, I'll show you.

>> Speaker 6: For tax reasons, they keep.

 

 

>> Paul Schmelzing: Right.

>> Speaker 6: The fireplace, for example.

>> Speaker 6: So that it does not count as a new house.

>> Paul Schmelzing: That's a good point. Okay, so absolutely it's a stylized way of-

>> Speaker 6: Just wanna understand the data.

>> Paul Schmelzing: Of dealing with quality changes, absolutely. And I believe Case Shiller would be the first to admit that there's so many quality factors.

 

Okay, but in any case, hopefully it's pretty straightforward to see that to splice this kind of survey data for the, for the first half of the entire sample with the much more robust modern data is problematic. If you simply ask homeowners what they think the value of their houses, that doesn't control for anything in many ways, okay?

 

That same kind of underlying-

>> Speaker 5: It's the same house. There's, there's a very large control.

>> Paul Schmelzing: Yeah, but I mean, if I simply tell you my own impression of what the house is worth, it might be very different from any transaction level data that is generated by actual sales process.

 

 

>> Speaker 5: I think the role of the rule that I've noticed in any people attack survey results that are self generated. But there's plenty of noise in it. But there's actually not much bias. So you ought to take what people say seriously.

>> Paul Schmelzing: Okay, I mean-

>> Speaker 5: The topic of conversation bar all the time.

 

 

>> Paul Schmelzing: Right, no, I see where you're coming from. I would absolutely think that the biases are multiples of the biases that are still existent in transaction level data because people have all these different motivations to deviate from the actual underlying price.

>> Speaker 5: Which tend to sum to zero anyway.

 

So good.

>> Paul Schmelzing: But I mean this data basis has been criticized by other economic historians, financial historians for some time now. I'm not the first to raise these questions over splicing these kind of things together. These data issues extend to all of these other advanced economies, including in that AER paper in the Knoll and that dealt with these other 13 advanced economies, okay?

 

For Germany, for instance, on which I will focus on exhibits in some of the next slides, they take one single city, namely Berlin for pre1930 and they splice together tax assessments, okay? And then for the post 1950 period, they use national statistic data that focuses on a much broader geographical.

 

So on this basis, I would immediately challenge the empirical robustness of a lot of these stylized long run secular housing facts out there that have been very prominent in recent years.

>> Speaker 7: To exploit data on rents, obviously there's asset price component to house prices as well, but presumably data on rents would also provide some cross-check on this and that should be easy to get going back in time.

 

 

>> Paul Schmelzing: To the extent we'll have time, I'll have one or two slides on rents and how they are constructed. My own impression is that it's actually harder to get rent data than the actual sales transaction data for various reasons. But good point on the rent part. Now the more holistic point here is that, and some of you might remember this from the previous interest rate results, it has been clearer and clearer.

 

I would posit that for all these econometric tests that we do on these time series in asset pricing, you get vastly more statistical power if you use robust long run time series, okay? And we made that point in the AR paper that came out just two, three months ago on the basis of real interest rates where the stylized fact that something big happened in the 1980s to global real interest rates disappears if you use longer time series, okay?

 

And housing with a depreciation rate of call it 1% or so. I mean there are papers that say if you have a well maintained house the depreciation rate is well below 1% per year. So housing is a long-lived asset, and for these Repeat sale indices especially, it should be very valuable to construct long run time series, okay?

 

Given that the longevity of housing as an individual asset. There have been leaps in the reconstruction over the most recent years in a lot of the associated variables that interact with housing as. I mentioned real interest rates, but this also goes for income growth. The national accounting literature has boomed over the last two or three years for the long run.

 

And we are really living in a new post medicine age on the empirical front when it comes to the reconstruction of national accounts, for instance. But also other financial variables, many of which I'm introducing in my forthcoming book over the very long run for advanced economies, for inflation rates, for risk premium and other assets.

 

So, it looks like a very promising exercise now to connect the new housing data to these other key variables that we know interact with this asset class over the very long run. So, the punchline of the real interest rate paper was that even 100 or 150 years of data, if you run Monte Carlo exercise and the like, give you very large false positive results.

 

All these standard economic exercises, whether it's the bipyron, structural break exercises, or stationarity exercises, etc. So it looks very promising to look at housing from that point of view as well.

>> Speaker 4: Treatment of depreciation. You said that if a house is well maintained, it's zero. Well, I think-

 

 

>> Paul Schmelzing: I didn't say zero.

>> Speaker 4: Let's say tax considerations. The capital depreciates, but then you spend money to reinvest and maintenance is kinda like an investment. So, when I hear the word depreciation, I'm thinking of before maintenance and before replacement costs.

>> Speaker 6: I agree with that cuz-

>> Speaker 4: Yeah.

 

 

>> Speaker 6: The reason that houses have longevity is because people keep making gross profits.

>> Paul Schmelzing: Correct, correct. No, my point was just that unlike other financial assets, it's not an asset that has a terminal value like a bond, for instance, that expires after 30 years, essentially. And this fact about the 1% was from a recent paper in real estate economics or so by a Dutch guy who tried to revise our understanding.

 

So the point is housing is a long lived asset compared to other assets.

>> Speaker 7: But there's a great deal of diversity within a country, the point Ken was making, my first job was in Pittsburgh and Pittsburgh housing went nowhere. And in California, housing prices have risen enormously. So what you're calling changes in the national level are radically different.

 

 

>> Paul Schmelzing: Yeah, yeah.

>> Speaker 7: I mean, does it make more sense to focus on major metropolitan areas or-

>> Speaker 6: You mentioned you're looking Berlin, for example.

>> Paul Schmelzing: Absolutely, yeah. No, I mean, we have one case Shiller index just for the 20 metropolitan areas, for instance. And absolutely that differs from the national index in many ways, okay?

 

Absolutely, and in the long run, these discrepancies multiply by big factors, obviously, okay?

>> Speaker 5: It depends what question we wanna ask. Yeah. Here's what we know about housing, it's a durable good. And the cost of housing in the long run is the cost of lumber and what it takes to put it together, plus the land and the permits, which are restricted, and the adjustment costs.

 

And Q on the up is bigger than Q on the down cuz most of it is about churn. People leave Detroit and go to San Francisco and the prices there don't fall as much as the prices here rise because it's an asymmetric adjustment cost thing. So, why are average house prices in the US high?

 

It's cuz there's churn, it's not that the, right? And there's all these Q less than one areas that are just quietly depreciating. That's a question of, is housing an asset class as opposed to the individual ownership of the house is your asset?

>> Paul Schmelzing: Yeah, certainly, I mean, these days you can invest in real estate through REITs and other listed instruments, etc, but absolutely it's often not considered an investable asset in the universe, okay?

 

But there's some debate, as far as I can see, on what extent can I get exposed to that aspect.

>> Speaker 5: There's a policy question, do our government thinks that the way to build generational wealth is to subsidize people to buy houses rather than to buy stocks? And a lot of us would say that's nuts, they should be buying stocks and renting houses.

 

So, there's a question, but of course, that question has to start with the heterogeneity issue that houses and even chillers seem to say that when house prices are high locally, they're gonna depreciate.

>> Paul Schmelzing: Right, yeah, we look a little bit at stationarity and persistence on the local level if we have time.

 

But that's a great point. At the moment, I wouldn't know how to have a section on the churn exactly.

>> Speaker 5: How to incorporate any nice exercise here, but obviously that's what's going on.

>> Speaker 6: I mean, you could do a decomposition if your data were rich enough. How much of the price increase is due to compositional effects versus just simply following one house the whole time?

 

 

>> Paul Schmelzing: Okay, I mean, I have a value weighted version of the index, for instance. Is that something that could address it?

>> Speaker 6: It's tricky.

>> Paul Schmelzing: Can I jump in here and this is related to some of the discussion.

>> Speaker 8: I'd like to hear an economic motivation for why we would look at the national housing price because this has already been mentioned.

 

It's a very distinctive type of asset for the individual who lives in the house. There's a great deal of hard to diversify risk within the relevant investment horizon of the individual. That's not present with the kinds of assets you talked about, at least not to the same extent in your earlier work.

 

So why is this even the right way to think about it? Another way to think about it would be what's the average try to characterize the risk that the average person who owns a house faces over a 10 or 20-year horizon. So I just wanna hear some economic-

 

 

>> Paul Schmelzing: That's one dimension that we will address. What is the risk for the individual homeowner to experience a fire sale event, for instance? The broader point is that as I will show, there's more and more evidence that we have these reach for yield dynamics in financial markets over space and time, including in 17th or 18th century Amsterdam.

 

There was a new paper in the RFS. So, even centuries ago we have this interaction between falling sovereign interest rates and investors switching from their portfolio from.

>> Speaker 8: Correct me if I'm wrong, I would think for the vast majority of homeowners that's not a consideration at all. They're gonna, Buy a house or they're gonna rent and if they buy, they're gonna be subject to a lot of risk between when they buy the house and whenever they want to sell the house.

 

 

>> Paul Schmelzing: So, one thing is-

>> Speaker 8: It's not about reaching for yield so much. Maybe it's reaching for a bigger house or something.

>> Paul Schmelzing: Right, I mean, one thing is that renting is a very old phenomenon. Even in the 15th, 16th century, we have a significant share in advanced economies of renters.

 

And they are buying housing as a product from people who own many individual homes, and they own it as investment.

>> Speaker 8: Okay, so for the homeowner class who are renting that I agree with that.

>> Paul Schmelzing: Right. So, maybe you're gonna tell me that most people were renters over most of this time period.

 

 

>> Speaker 8: Is that the case?

>> Paul Schmelzing: It's close to 50% according to some estimates.

>> Speaker 8: Looking back over centuries.

>> Paul Schmelzing: Over centuries?

>> John RYLOE: When we think of the long run returns of housing, it's got to be the rent. The rental value of living in there is longer term. The long run price is got to be the marginal cost of building it or the marginal cost of rebuilding it constantly.

 

In the same way, there's a risk of owning housing, but there's also great insurance of owning housing that you are insured against the rent going up.

>> Speaker 7: Yep, correct.

>> Paul Schmelzing: So, we look at rents in particular in a separate slide. If I may, I will direct your attention to this slide here which tries to make the point that this is also an amazing point in time, methodologically to rebuild these new long run housing indices because of major leaps in the digitization and really the archival side.

 

What you see here is a project from the city of Nuremberg, which was one of the key financial hubs of early modern Europe and the world. Essentially the home of the Fugger bank and others. They have just launched a new project called Topo End, which is a housing unit level reconstruction of every individual house in that city as far as the records permit, in terms of all the information that is associated in the cadasta records and others, including the ownership of course, including the any transactions, including any inheritance, including any fire events, including any mortgages, including any insurance events that are associated with the housing.

 

So, this individual city has done that for 4,000 houses already. They are adding a couple of hundred every year or so. The amazing thing is that if you now go to Google Street View or so and look for these individual unit level descriptions and records, many of them still exist, okay?

 

A vast share essentially, okay? So, in this case for the Bagara plus number 10, okay? This is a house that is in continuous existence, at least you know the address here, okay? For over 600 years, essentially, okay? And this exists for other cities. This is done for, call it dozens cities in recent years.

 

And this allows you in some ways really to build that exact geographically weighted national index that focuses on these large metropolitan areas and tracks the sales values and all these other relevant events. Yeah, please.

>> Speaker 4: You have to worry at all about the sort of selection on survival.

 

I mean, similar issues come up with prices, CPI more generally, right? That the houses that survive, maybe these are in less dynamic parts of the city where less rebuilding.

>> Paul Schmelzing: Correct.

>> Speaker 4: Quality change might be smaller than city as a whole.

>> Paul Schmelzing: Yeah, absolutely. Let me get back to this.

 

For the Nuremberg sample, for instance, the amazing thing is that we have up to 490 years of individual houses of continuous records with all the individual events that I mentioned. The average is over 220 years for each individual property that we can continuously track. Okay, these are the new digitization efforts.

 

Yeah, please?

>> Speaker 6: I recall when I went to Nuremberg, I thought that there was quite a bit of bomb damage from World War II and Nuremberg was one of the cities that decided to build the historic part where, say, Frankfurt just decided to start new. So, what do you do with houses that are just reconstructed as replicas of what was there before?

 

 

>> Paul Schmelzing: So, that's true. Nuremberg was damaged, as other German cities. Parts of the city were survived. And I believe this part of the town, the Altstadt, with these 4,000 units, they escaped World War II relatively well compared to other areas. But absolutely, we have this survivorship bias in the data, right?

 

And we talked about fire events and not just World War II, but over, over history. The 30 years war perhaps was the most destructive material event for housing. We'll come to that. There are various ways of dealing with these events. So, one way is to just kick any house out of the sample that has one of these major events and focus on the individual houses that are continuously tracked.

 

But actually, we look at this in the context of the fire sale frequency. But the other ways is to try and deal with the perspective of an investor, for instance. And we will see that we have these dynamics where distressed investors come in after these events, they buy the destroyed house, they rebuild it, and they make a lot of money in these transactions.

 

And this is going on on a scale that makes you hesitant to just ignore the phenomenon, okay? And so in most of the benchmark series that I will show you, these events are included, but there are alternative versions where we exclude these kind of extreme events. So, this is an amazing time to rebuild, just through these new digitization efforts, to rebuild series like this, okay?

 

I shall mention that I brought one of these older compendium that I also draw on. So, in Germany, we had this phase called the historical school in the 19th century or so, where the precursor to these digitization eras existed, people wrote books like this with just all the catastrophe data from major cities.

 

This is for the city of Munich. This is just one of six volumes for the old part of Munich. You find for each individual address, you find the ownership history, you find transaction data. You find, if there was a major fire event, if the owners bought insurance for the house, and what level that insurance was, if they took out a mortgage for the purchase of the home.

 

So, I merged these new digital databases with older material, and this really creates a comprehensive. And here you see, for example, for Berlin and Munich, these kind of volumes have existed for a longer time. Okay, so we also observe, importantly, foreclosure sales, okay? In Latin, these events are called Subasta, when an owner is forced into a foreclosure, okay?

 

And fortunately, most of these databases and older records, they specifically Specifically because this was an event that had to be legally documented if there was a foreclosure sale, and we can compare essentially these events with some recent efforts to do this for the US great, okay. So let me stress that I focus on Germany for a number of these exercises now, but this is more of an exhibit for other advanced economies, okay.

 

So in the Knoll Schulich paper and others, they show very nicely the correlation of German markets with others. Now, this doesn't necessarily mean that there's these correlations, of course, but the basic methodology can be replicated for other markets and for a significant share of financial market activity over the long run, I would say, okay.

 

And so, excuse me, at this stage, it's not yet a big data exercise or anything, it's more about the conceptual approach, okay, so I would really appreciate, especially on this kind of early stage, more fundamental pushback is super helpful.

>> Speaker 4: Now there's all sorts of questions to answer with this.

 

And back and forth and the flow and when it goes up and when it goes down. But for the question of long run return, I want to emphasize the survivor bias is a big problem. If I can send you back in time to 1930 and you only know one thing, the name of five companies that are still alive today, you could make a fortune in the stock market and people have done this.

 

That's all you need to know is that they're alive then and they're alive now, whole cities are not alive anymore, so, that the house lasted long enough to be sold is a survivor bias. So as far as the long run average turn question, that seems to be a big problem, 100 other questions that are could be fantastic in this data, right.

 

 

>> Paul Schmelzing: Yeah, I wonder what the mutual fund managers, If I could just tell you who's in business right now and go back and buy them 10 years from you go, you'd make a fortune, right. I mean, we have a decent share of these early modern investors, they obviously held properties in different cities trying to diversify that exact risk, right.

 

And so the index that I'm showing you in a few slides is basically taking that diversified risk that you would get with a nationwide exposure. You don't know which individual city will survive or be destroyed by World War II, but, this tries to tell you what the diversified exposure brings you, just you would invest in a diversified ETF or so, right.

 

The data is based on something that you knew was alive at the end, so, yeah, you need I mean, to do average returns, you need to know everybody who's alive at the beginning, including the ones that died. Yeah, correct, I mean this is the same issues that plague stock market indices that people have reconstructed.

 

 

>> Speaker 4: But, maybe you can correct, you kind of know how many cities disappear.

>> Paul Schmelzing: Yeah.

>> Speaker 5: It seems the things you can do in the data, right, because you have all the houses historically so you can at least create balance tables on the values historically and then pre trends.

 

And that will sort of tell you something about the houses that survived versus those destroyed today, right if they look similar back in 1930 in terms of levels and trends, right. You could do some things to, try to address that, these concerns, right?

>> Paul Schmelzing: Yeah, I made a note and exclamation mark here in my notes to think about this more.

 

It's an approach that is I said, used in these long run equity indices as well, where you don't know which individual companies exposed, right, but so that's well taken. Okay, so we have all of this, first of all, we have all of this transaction level data for the individual houses, I ignore everything that's not an actual sale for the moment, okay.

 

So we have the information, when is the house inherited by a relative, when is there a transfer of house with other person or so let's focus on the, the actual sales. We have hundreds of observations already and a couple of just interesting stylized facts is that there seems to be just high level, there seems to be a long run decline in the holding period of the individual house consistent with what we're seeing in modern US Markets.

 

For instance, where we're talking about a holding period of 8 years or so for the individual residential property, on average, between sales and that number, we start in the mid-30s or so holding periods, there's a secular downward trend and that suggests these markets become more liquid and et cetera.

 

And there's a corresponding long run upward trend in the average gain that a seller of the residential property makes okay from something in the, in the low 30s or so to something more like 55 or so percent in the modern age, okay.

>> Speaker 8: If we, piece these data points together and so these general figures are consistent with modern data, I would say this is essentially the blue line, how the arithmetically weighted index now looks.

 

 

>> Paul Schmelzing: Okay, If I piece it together, these 11 cities just based on actual sales transactions, okay, you can create a value weighted index that I can show you different ways of just focusing on larger cities, smaller. This is a sample that tries to be representative of the national index, just the Case Shiller Index tries to be representative of US Trends, okay.

 

And then what we do here is we piece that together with the more recent data in Knoll Solaric and these other very recent papers that try to reconstruct the last 100 years or so, 150 years, and basically claimed that nothing much happened prior to 1970, okay, that is in red.

 

And so. Here the punchline really is that actually there's a lot of action in the early modern period, certainly prior to the 1970s, okay, in real terms, residential house prices go up by a factor of 10 to 15 over these centuries. They go up by a factor of 3.5 alone between the Congress of Vienna in 1820 and World War I, okay.

 

And that suggests that actually this idea that we had these stagnant, pretty boring global housing markets prior to the 1970s or so is a feature of the particular data sample that we have constructed so far.

>> Speaker 7: Do you have a log scale, I was wondering how far off exponential is.

 

 

>> Paul Schmelzing: Yeah, we also have lock terms. Yeah.

>> Speaker 5: It looks like it's not that far off exponential, right, but a constant, growth.

>> Paul Schmelzing: Right so this is, yeah, we have locker data as well, but this is just showing it's growing exponentially visually. Yeah. No, these are two different axes, you see that, right, so, but, yeah, so this is.

 

Just way too big to be believed.

>> Speaker 4: 600 times more expensive house now than in 1450. So let me start with this has got to be nominal, right?

>> Speaker 7: Real, this is real. So a house-

>> John RYLOE: How's the price in Mexico?

>> Speaker 7: So certainly, the quality of a house now is better than 1450 though I don't know.

 

 

>> Paul Schmelzing: Yeah.

>> John RYLOE: Yeah, looks pretty nice to me still. Although now they have toilets that are. And air conditioning. And air conditioning. The marginal cost of building a house has if anything gone down. We have power tools now, and the marginal cost of maintaining a house has gone down.

 

If you want to put a stone block on top of a stone block, we haven't seen to do that now. So it's got to be much cheaper and I guess wages are high.

>> Speaker 6: Not to mention the inputs Dan Stitchel's thing about the nails. So when a house would burn down before everybody would run out to pick up the nails because they were so expensive, they were done by hand.

 

 

>> John RYLOE: Not expensive in near terms cuz wages hiring people to make nails by hand.

>> Paul Schmelzing: Yeah, there was a great paper on nail prices on the very long run.

>> John RYLOE: So just tell me what's wrong cuz factor of 600 memory, does it start at 100 or does it start at 1?

 

 

>> Paul Schmelzing: All right, at 100.

>> John RYLOE: Starts at 100. So houses are 600 times in realtor more expensive now than they used to be.

>> Speaker 5: Not 60 quantity control now to be clear, just- 100, right, you said.

>> Paul Schmelzing: The left hand side starts in 1465 is index 200.

>> John RYLOE: Okay then, so 100 does 6,000.

 

 

>> Paul Schmelzing: To emphasize again, this is not quality controlled. Just like very recent indices in the AR paper are not quality controlled. Okay, it helped a lot. And on this basis the stylized claim, is that nothing much happened. And it's size controlled also, right? So when you say size control-

 

 

>> Speaker 7: When you're comparing then a one bedroom hut in 1450 with a dirt floor to correct what's there now which is a 5,000 foot Meg mansion.

>> Paul Schmelzing: If that's what you mean by size control, yes, it's size controlled.

>> Speaker 7: What's happening with rear weight?

>> Speaker 6: You were doing repeat purchases of the same house.

 

 

>> Paul Schmelzing: Yeah, in the appendix I'm showing how the average size of these houses changed over time. And we have how the floor space changed, correct.

>> Speaker 6: So this is the remodel.

>> Paul Schmelzing: Yeah, I mean if you add another floor for instance into the same address.

>> Speaker 6: With central heating.

 

 

>> Paul Schmelzing: Correct, we need heating under constraint. One big factor I will emphasize is how we deal with big inflation shocks over the long run, okay? The approach in this Knoll Shurek paper for global housing markets was just to exclude big inflation shocks like the German hyperinflation, or it happens in other countries, okay?

 

It turns out to make a massive difference for that index level. How we treat the 30 years war, for instance, which was a massive inflation shock. And in the actual paper I show you how the index looks. If we include that massive inflation shocks, we're going down from a factor of 600 or so to a factor of 5 to 6.

 

 

>> Speaker 5: Okay.

>> Paul Schmelzing: It makes a massive 600.

>> Speaker 9: You aggregate these repeat sales transactions. So are you just taking the mean or you look at the mean-

>> Paul Schmelzing: Yeah, this is arithmetically weighted.

>> Speaker 9: So I worry about extreme values like John's example. If you go from the one room with a dirt floor to somebody builds whatever the latest version of the house is at that time period.

 

So it'd be useful to know whether it looks materially different if you were linked to get base this on median values rather than means. That's a good point too. That would be easy to do. Catch up, falling behind.

>> Paul Schmelzing: Okay.

>> John RYLOE: So how is the CPI constructed or whatever, to get into real terms?

 

I mean, and why does it not work during hyperinflation? I would think that if you're doing things in real terms, you would wanna include major price increases and housing prices should be increasing also.

>> Paul Schmelzing: No, I would absolutely agree with you. But to be consistent with these existing indices, this version of the blue line excludes these massive inflation shocks.

 

In the paper I'm showing how it looks with the inflation shocks included, okay? Just interpolate across the inflationary episode. Correct, yeah. Or put it at zero. Okay, so I need to run a little bit. So the one interesting application is just that we have these new studies on fire sales and the 0809 crash as a fire sale event, what extent is that comparable over space and time?

 

And in that sense on the individual housing unit, how does that risk change of an individual fire sale? And how are these spikes in fire sales on the aggregate level connected to macro shocks or what drives these shocks in housing as an asset class? So on here on the chart you see one definition that follows John Campbell and co-authors who've written fire sales just taking sales transactions with more than a 25% discount over the previous sales process price and defines that as a fire sale event, okay?

 

And that's consistent with US housing data over the last 30 years or so, where the average fire sale event produces 27% discount based on John Campbell's.

>> Speaker 9: What's the definition of sale? Fire sale, that's two rather different meanings in American English.

>> Paul Schmelzing: Okay.

>> Paul Schmelzing: To be clear, it's not necessarily an actual fire in house that we're talking lot.

 

But it can be driven by any particular, correct.

>> Speaker 8: The large decline 20, 30% decline in a short period.

>> Paul Schmelzing: 25%.

>> Speaker 8: 25%.

>> Paul Schmelzing: Yeah, correct. So across the entire sample, what is the kind of frequency we're talking about and how does this evolve over the long run?

 

Correct, and so we have these big spikes in fire sales that historically, they are associated with big geopolitical shocks. Actually, you see, the first big, excuse me, it's not in. We could call them buying opportunities. We have the first big spike in fire sales in the late 15th century.

 

It's connected to a major regional war in the south of Germany where Nuremberg and other cities battle each other, burn each other's properties and we get this massive spike in fire sales up to a level, interestingly, that is closely comparable to the spike in Massachusetts fire sales around 08 and 09, okay?

 

Where Campbell et al, they measure 28.5% fire sales based on a closely definition. In a war where they actually burned down houses, I would have expected the price of the remaining ones to go up.

>> Speaker 6: Yeah.

>> Paul Schmelzing: Assuming the San Francisco earthquake is famous for that, thanks to Milton Friedman, I guess.

 

 

>> John RYLOE: On the other hand, if they burn down half the house, that means the army's there ready to burn down the other half. So you better get out of town.

>> Paul Schmelzing: Yeah, that's a good point. I mean, something for me to look up, I guess. So the next big spike is the 30 years war.

 

It's probably, based on my data, the biggest housing shop ever. In terms of the data, in terms of the material destruction that we're talking about, there's nothing that comes close, even World War II or so. And so again though, the peak in the fire sales based on that definition that tries to replicate the Campbell definition is actually very close to 08 and 09 left they using Massachusetts data, only.

 

 

>> Speaker 5: 08 and 09 in what century?

>> Paul Schmelzing: Say that again.

>> Speaker 5: 08 and 09.

>> Paul Schmelzing: Yeah, I mean they only.

>> Speaker 9: Fire sales have gone to zero.

>> Paul Schmelzing: In normal times they're much lower, of course.

>> Speaker 6: But not in Arizona.

>> Speaker 9: Are we talking about houses bringing up or.

 

 

>> Speaker 6: No.

>> John RYLOE: This is Vegas.

>> Speaker 9: Not a literal class vocabulary problem. This is how often an individual house in your sample gets sold for more than 30% less than it was bought 25% more than it was that. That the definition of a virus, Correct. In real terms, in nominal terms, for me, it'd be more informative just to show a time series of box plots of the price changes rather than just pick some arbitrary threshold.

 

You have to average over five year periods or something, right. Fire sales might also have booms the same time if it's from the box plots, I mean, everybody needs to leave Nuremberg and go to Wittenberg. Then we're gonna see those go down and those go up, after the 30Years War, then, however, we declined to a level of, call it 5 to 6% in normal times, okay.

 

 

>> Paul Schmelzing: We get these occasional spikes of the Napoleonic wars, for instance, but we seem to have this big secular decline in bios, okay for the individual housing level. And interestingly, again, this level, call it five and a half, 6% or so, is closely comparable to the level that Campbell measured in their paper for normal times in US housing markets, based on Massachusetts data.

 

So, but, so you're associating worst, so this was not a boom that led to going back to normal prices. This is associated with something horrible, but it would be lovely to have just some stories, why was this house sold? Right. So what I mentioned in the, in the underlying sources, we have explicitly foreclosure events, they're closely tracking, this definition, okay.

 

And so it's not the price boomed and the price went, it's just people's businesses were destroyed and so they couldn't correct it. Kind of all sorts of reasons, we don't, unfortunately, we don't know in these sources what they don't tell us why was there a foreclosure event? Was this guy, did he, did he speculated on the stock market or what happened that unfortunately you can't tell, but, the general prices ramp up and then fall, or were general prices just hanging around and just more people.

 

So index it relative to the, index that on the national level that we construct and say, if there's a 25% deviation to that, to that index level, then we call it the fire sale. So that's one alternative here, I just tried to replicate the methodology in some of the inferential papers.

 

So again, this peak level and also this average level look remarkablye something we see in the data for recent decades. Okay, that's very interesting because to some extent this narrative is that housing markets modernized around the year 1700, right? They took their modern shape in terms of the individual housing level risk, and this is consistent with a lot of the, what historians have dug up, on the city level, the firefighting emerges around the year 1700.

 

I mean, the London fire in the 1666, etc, the last major fire event in an advanced city, afterwards there are massive reforms, how to deal with the fire risk in cities, okay. There's a wonderful book by a historian, Cornell Swirlin, who focused on the Prometheus Unbound, who looked at fire incidences in cities over centuries.

 

That's very consistent with the picture that around 1700 there are so many reforms in terms of you have to keep a gap between individual houses so that in a fire event it doesn't spill over as easily to the next house, etc. So that is consistent, I would say, with this idea that we have this massive decline in individual.

 

These might be real buyer sales. So again, yeah. Buyer goes and I still owe the mortgage, so foreclosing the mortgage and the bank gets to get this, incorrect. So I should be clearer here on.

>> Speaker 6: Distress sales.

>> Paul Schmelzing: Distress sales.

>> Paul Schmelzing: Okay, so let me focus on this first, so the interesting thing is that Robert Schiller, many others, directly connect their index level with the behavior of interest rates, right?

 

And that's an obvious connection to study because here, this is from his book here, excuse me. The black line is US Normal interest rates of the same horizon,okay as we see this massive acceleration in US Real house prices, we see this decline in interest rates, okay? And by no means, Bob Schiller is by no means the only one to posit this close connection between the behavior of interest rates and mortgage rates in particular, and the level of house price.

 

So this credit channel suggests that falling interest rates, falling mortgage rates should pump up house prices.

>> John RYLOE: Why are you doing nominal Interest rates and talking about a credit channel, price over rent equals 1 over R minus G, so I think just the real interest rate goes down, the price of the long asset goes up.

 

 

>> Paul Schmelzing: This is the version he shows, we can also do it in real interest rates.

>> John RYLOE: Yeah.

>> Paul Schmelzing: I'll show you in a second.

>> Speaker 9: You seem awfully wedded to repeating the mistakes of everyone in your literature.

>> Paul Schmelzing: No, this is probably a chart many people know because they've read Bob Schiller's book and it's just everybody you cite, they're doing something terribly wrong.

 

 

>> Speaker 7: You feel you have to do it wrong the same way.

>> Paul Schmelzing: No, that's fair. But it was the natural point of departure then to link these house price trends with some of the corresponding work on real interest rate over space and time, right. Because as I mentioned, there was this beautiful paper just last year in the Review of Financial Studies on Reach for Yield, which suggested that in Dutch housing markets, when sovereign interest rates fell, people actively rebalanced their portfolio and shifted from sovereign bonds into housing.

 

Okay, if that's true, and the author is quite convincing, there's an example where I would say. We get it in finance, we can individually shift our portfolios, but collectively we got to hold with the market.

>> Speaker 8: Yes. So, yeah. who bought those Dutch bonds then?

>> Paul Schmelzing: The change on prices has nothing to do with the change on quantities, yeah, somebody held the government bonds.

 

Right so they all tried to move into housing and in doing so they pushed the house price up. Right.

>> Speaker 8: Now I find that principle recognized in the financial press.

>> Paul Schmelzing: Just amazing, isn't it? Every buyer there is. That's we're done. But another example here is the JPE paper the Giustiniano et al who studied the 0809 developments and the credit channel of the the housing boom and bust in the USA.

 

Also claimed that this credit channel had really one of the most important channels was through falling real mortgage rates in the US since the 1980s, okay. That naturally led to the question is it really something that emerged in 1980s, and what is the exact connection to long run trends in real interest.

 

 

>> Speaker 10: Is there anything interesting about demography, so one thing that's been happening in the US shift towards smaller family size, right, which increases average demand for space.

>> And you have really long data with huge demographic changes, so maybe you can provide some really interesting evidence.

>> Paul Schmelzing: So in the paper, in the current draft I say more about the population, it comes up as statistically significant in a lot of the regressions.

 

And there's a strong literature on baby boomers, how they pushed up house price, etc., and I think that's another promising variable. The whole credit story is stronger though in both in the data and also in terms of the long run critical picture that we have for now. But I would by no means diminish the role of the.

 

 

>> Speaker 9: I'm a little, so the buzzwordy, passing along these buzzwords. Lower interest rate is not a loosening of credit constraints, it's just a lowering of the price of credit. There is something else where people think there's a non-price rationing of credit and that's some other phenomenon. But noticing that the interest rate goes down and the price goes up doesn't have anything to do with reaching for yield or credit constraints or stuff of the sort.

 

It is not your fault, it's just people write this stuff, even Wall Street Journal.

>> Paul Schmelzing: So what I did mention is that these sources also tell us something about the mortgages that we're taking out in each individual transaction, right? And one thing we can look at is the loan to value over space and time and here we don't see any, at least to the extent I can reconstruct it at the moment we don't see anything, any secular trend, okay.

 

It seems to be more stable than many of these other variables, it is not available for all the cities and not as comprehensive as I would like. But this is one of the things that we can use to test this constraint idea, right, and the-

>> John RYLOE: This linking of prices to rel.

 

It's kind of a sense higher prices, lower real interest rates and you kind of get this trend from 1980. But the boom and the bust and the boom again seem to have nothing to do with the level of real interest rates.

>> Paul Schmelzing: Yeah, I mean I'm-

>> John RYLOE: Real interest rates went down from 2008 to 2016.

 

 

>> Paul Schmelzing: Yeah, you could say this was financial.

>> John RYLOE: There's always supply and demand.

>> Paul Schmelzing: So as a reminder, this is how sovereign real interest rates look like for advanced economies. We call it the global level, this is from my forthcoming book and from a paper this year. And one of the big punchlines here was that there's something big happening in sovereign real interest rates in the 16th century.

 

Much bigger in all these standard tests than 1914 or 1918. This date by the late 16th century shows the strongest and really across all the countries and all the plausible inflation constructions that you can use, etc., something big happened in sovereign real interest rates in the late 16th century.

 

We can do the same for real mortgage rates. Okay, so let me show you-

>> Speaker 5: What country, is this the average for world?

>> Paul Schmelzing: This is for eight advanced economies, including when the US comes into the sample in 1700 when they are developed.

>> Speaker 5: But US and UK have been historically much safer than some of these other advanced countries.

 

 

>> Paul Schmelzing: Correct, I don't have the time to go into the construction details, but absolutely using the late 16th century. Okay, but it looks like it's the entire 15th and 16th centuries where there's this big downward trend in your chart.

>> Speaker 6: So is that what you mean by bit, you kept saying something big happened, but I couldn't.

 

 

>> Paul Schmelzing: Yeah, I mean we get the trend station downward down. Correct, so what you see here is if there's one asset that's older than sovereign bonds, it's actually mortgages. Okay, so here you have an example from a mortgage contract from the year 1624, okay. April 30th, between a debtor and a creditor, here's the monastery on the creditor side, lending to this.

 

I believe she's a widow buying a new house at 5%, a mortgage contract. This is part of a sample of the series I introduced in the book, but we can use it here to reconstruct long run real interest real mortgage series over space and time, okay. And it covers the same city sample that we use for the house price index, and it allows us to say hopefully something more about this idea of a interaction with interest rates and credit.

 

Here you see real mortgage rates between the year 1675 and 2024, okay, for Germany. And like real sovereign interest rates, it turns out that this is a trend stationary series with very few structural breaks. If you use all these conventional tools that people use and the big inflection point, however, occurs roughly 100 years after the big inflection point in sovereign rates.

 

So here specifically, so to be clear in this, the Knoll Schurich paper, they use hyperon structural break tests to make these claims and to underpin it. I can use the same test for these much longer series where presumably we have much more specific power. And it turns out in mortgage markets, something big happened after the 30 years war, okay, the late 17th century.

 

I mean the structural break test, you're fine. I'd personally be more interested just to see a well annotated chart of major economic and geopolitical events that happened. Okay, so I could put that together with what I see in the picture.

>> John RYLOE: Yeah. Going back before the 20th century, my knowledge of history is pretty hazy.

 

 

>> Paul Schmelzing: No, that's a fair point. I mean here, the spike, for instance can I show it the big spike here in the samples, Napoleonic Wars. Okay, so when you have risk events, people go into sovereign assets and that kind of thing. No, that's a fair point, so the next big one is for instance, the interwar period when you had a lot of volatility.

 

In Germany, obviously, and-

>> Speaker 5: Mortgages, non-recourse or?

>> Paul Schmelzing: Say that again?

>> Speaker 5: Are they non-recourse loans, or?

>> Paul Schmelzing: What do you mean exactly?

>> Speaker 5: Well I think in the US if I walk away from my house, the bank can't take my other assets.

>> Paul Schmelzing: Okay.

>> John RYLOE: What state are you talking about?

 

 

>> Speaker 5: In a mortgage-

>> Paul Schmelzing: Any state.

>> Speaker 5: The assets. Of course they can, they don't.

>> Paul Schmelzing: Even when they can, they don't, because you don't have any other assets.

>> Speaker 5: But in Germany, if I walk away from my house, can I get other assets like my pension?

>> Paul Schmelzing: Put you in prison.

 

 

>> Speaker 5: Yeah.

>> Speaker 6: Better wrap up.

>> Paul Schmelzing: Yep, so I will add the big geopolitical events here for sure, that's a good point.

>> Speaker 8: Is the red an ex-post rate, or what is that?

>> Paul Schmelzing: It uses the same, it's a seven year lacked in realized inflation approach following what we did for sovereign real interest rates.

 

A lot of that volatility is just inflation volatility because no one's signing mortgages at a negative 8% nominal rate. Absolutely, in the interest of time, let me skip to one more thing I think that's interesting. So, this is another great statement from Bob Shiller's book in which he raises the question about the relationship between house price growth and income growth, okay?

 

And he says in his famous book here that there is very little on the theoretical side or empirical side that should make us assume that house price growth can outstrip income growth for any sustained period of time, okay? Again, based on this early part of the sample where we have the stagnant house price index in the US and many other countries.

 

I mentioned that we have made these tremendous inroads into national accounting in recent years for both US, but all these other post medicine papers and projects. This allows us now to take a closer look at this idea that there shouldn't be any close premium of house price growth over per capita income.

 

So just using these latest data series. So here, you see it for the US first just indexing real per capita GP and the median house sales price for the US at 1997, when Shiller and others say this is when the housing boom started in earnest, okay? The blue line, as you can see, tends to outstrip the green line.

 

In other words, house price growth tends actually for the last almost 30 years now tends to do better than income growth, okay? So even for US data, it might make you question this, just the robustness of that statement, to say, it can't deviate all too long, for the last 30 years it seems it has deviated quite a bit.

 

 

>> Speaker 6: So how does this relate to Chad Roy Wright, given co-authors Neil Kindlebergian analysis?

>> Paul Schmelzing: I would need to refresh my knowledge on that.

>> Speaker 6: Because they claim that there was actually an underlying trend for house price increases that was based on fundamentals, is that you had, they didn't say irrational exuberance, but basically overshooting.

 

But then that's why it started going back up again afterwards.

>> Paul Schmelzing: Okay, okay, I would need to revisit that.

>> Speaker 6: Yeah, but I mean for the US, I mean that's an important paper.

>> Paul Schmelzing: Okay, let me catch up with you, perhaps after and check that out but it's quite a persistent phenomenon in any way and it has in any case intensified over the most recent years, okay?

 

Two different eyeballs, these look like very stationary, those two look like they're trending together to me, we're exactly right back now to where we were in.

>> Speaker 6: When is that?

>> Paul Schmelzing: 2020, and back to where we.

>> John RYLOE: The starting point?

>> Paul Schmelzing: Because that's, if you drop the line.

 

 

>> John RYLOE: 1997.

>> Paul Schmelzing: I mentioned, we let it begin in 1997 because that's the claim that the housing boom started in around 1997. You can also choose one of the adjacent years or so. It doesn't make a big difference, but we can, we can now use this long run data to analyze the relationship between the two.

 

And it turns out that for most of history, certainly house price growth was above real per capita income growth, okay? In fact, all the way to 1910, the early 20th century, house price growth was meaningfully above real per capita income. So you say people are devoting a steadily larger share of their income to housing?

 

Can't be under zero, can't be over 100.

>> Speaker 6: Well, how can they do if they're, they're already devoted higher shares to medical care, right?

>> Paul Schmelzing: Yeah, for instance.

>> Speaker 6: Paul Jones

>> Paul Schmelzing: So, yeah, I mean, we don't have the time unfortunately, to talk about wealth inequality and how it connects to this phenomenon.

 

But I would say it's consistent with some of the new top income share reconstructions.

>> John RYLOE: Each individual has a steady fraction of their income devoted to housing, just the houses are owned by richer people. Right, that's one explanation.

>> Speaker 6: But you said that approximately 50% have owned homes throughout the entire time, which was sort of the fact that every time I think that thought, I'm reminded of that fact.

 

 

>> Paul Schmelzing: I mean, one thing we don't adjust for enough probably is the movement from rural areas into the towns, right? Yeah, over space and time, which is a big factor. The urbanization trend itself is big over these centuries that we're talking.

>> Speaker 6: Which in a sense Ken's original point.

 

Exactly, which is, is it land or is it housing?

>> Paul Schmelzing: But using all these new data sets, widely cited, it turns out that this particular period, 1963 to 2020, is more of an outlier perhaps than we appreciate, okay. Just if we take this period, this is when these structural break tests in the AER paper, they claim this is when the housing boom started their samples.

 

So it might be a function of this particular construction approach, I'm arguing that made this idea so prevalent that income growth should outstrip house price growth. If you take virtually any other period and sub period in history, the opposite is true, including the last 30 years, okay. And so that should perhaps make us question this sacred assumption that there has to be this theoretical and empirical regularity that income growth has to outstrip real house price growth.

 

The opposite is true for most of history, okay. In that sense, maybe the past 30 years or so were not so much.

>> John RYLOE: Theory was that they have to be the same. The fraction, I guess if it's a normal growth, it's probably a income elasticity of one. So that should be the ratio, I mean, you can't get too far out.

 

 

>> Paul Schmelzing: Yeah, yeah.

>> John RYLOE: We can't get a factor of 60 grow in house price growth if income growth is exactly the same. Yeah, like I said, I think.

>> Speaker 6: And if everybody's still buying, half people are buying.

>> Paul Schmelzing: I want to dig into the exact movements from rural to urban areas much more and analyze household budget shares and what that fraction is over space and time.

 

Okay, that's an important extension.

>> Ken Judd: Yeah, no, I mean, if people are moving to cities and the cities, and the land in the cities is constrained, but the jobs are good. Then they're, they're consuming, you know, a third of their income as, as everybody else is, you know, but their incomes are higher, housing prices are higher.

 

And you're gonna measure that per capita the housing prices are up relative to per capita income, but it's all, it's, right.

>> Speaker 6: Normalized.

>> Ken Judd: People in the country lived in houses too. Their housing prices aren't going anywhere, and their income is not going anywhere.

>> Paul Schmelzing: Okay.

>> Ken Judd: I mean, there's a discipline cuz everybody spends 30% of their income on housing, that really gives a discipline to these numbers.

 

I would kinda use it the other way around to help you discipline long-term trends. If your long term trend says I'm paying 99, 99.99, 99.999% of my income on housing or 520% of my income on housing, something's wrong with the house price index.

>> Speaker 6: The other thing is, with the historical data, you have to make sure that the implicit income generated by living in your own house is included in the GDP numbers.

 

 

>> Ken Judd: The owner occupied?

>> Speaker 6: Yes.

>> Ken Judd: Yeah. Well, the rate of return number too.

>> Speaker 6: Yeah.

>> Ken Judd: So the big problem, the rate of return number is you get the rent, which you got to pay for the maintenance and the taxes.

>> Speaker 6: Yeah, but if you're excluding that.

 

 

>> Paul Schmelzing: I will need to double check how they account for these factors.

>> Speaker 6: Yeah, but probably better than the historical data.

>> Ken Judd: Hardly.

>> Paul Schmelzing: Okay, so the title of the paper is alluding to the emergence of the safe asset, of course. And here I try to fuse some of the insights from other projects with this new insight from one of the largest items on the capital stock, the housing, okay?

 

And so we've seen that after 1700 or so the fire sale risk at least suggests that, you know, there was not another major leg down in the individual housing risk. While at the same time for sovereign assets and for the public sector, we all the measures that we can plausibly use suggest a continued secular decline in risk, fault risk, inflation, volatility, for instance.

 

And trying to put the pieces together in that part of the project, comparing the risk profiles of the public and the private sector, okay? And making sense of the fact that the spread between private and public assets from around the year 1700 seems to secularly increase. In other words, the risk premium in the private sector relative to the public sector secularly rises from that point in time.

 

At least that's what these tests suggest and what also the historical narratives underpin. So this is consistent with a couple of other projects that I've worked on, what you see here is bank distress risk over centuries. This is from a project with Andrew Metric where we measure bank distress over the past four centuries.

 

Which suggests an increase in private sector risk in the frequency of bank distress, while we have at the same time a decline in all these public sector risk measures that we can use, okay? And so these two secular trends here you see the spread between public assets and private assets for Germany, starting in the year 1465, we have this long run downward trend.

 

But starting around here you see starting around the mid 17th century, late 17th century, when we have both the inflection in the real mortgage rates and in these indices that I mentioned. When this spread between public and private assets secular declines, or if you want to look at the reverse, the risk premium for private sector secularly starts rising, okay?

 

And this contextualizes this observation that in recent decades equity risk premia have stayed pretty high while the safe interest rate has continued to decline. Here the tentative suggestion is that we are also dealing with a very long run phenomenon that started way before the 1970s. So if we try to measure these two pieces comprehensively, we see it in French data that we can reconstruct this from my book.

 

We see it in US Data here on the right hand side panel, uses US mortgage rates versus US Long maturity government assets, public sector assets. We have actually if you look at this long run decline in public sector risk premia, or if you want to look at the reverse, the top prior spread, okay?

 

This consists of the data that we see in all these other advanced economies for the secular term, okay? And so this decline, this pool is being driven, I suggest, by the public sector becoming safer and safer and safer. While we have relatively stable or moderately increasing risk from these different variables in the private sector.

 

Including from the fire sale risk that I showed for housing, moderately increasing bank distress risk in the private sector. And so this part of the paper tries to holistically look at the risk profiles of these two sectors over space and time using that new big item in the capital stamp over centuries, okay?

 

So again, the message is very much that I use Germany as an exhibit, but it's really consistent with all these other countries where we have new data. And where we once more have this punchline that perhaps nothing that special in the 1980s happened, after all. If we look at these longer time series that we can now test for all these different variables, so thank you very much for all the great comments and pushback.