The rate of return on everything

There is a new NBER paper on that topic by Òscar Jordà, Katharina Knoll, Dmitry Kuvshinov, Moritz Schularick, and Alan M. Taylor, here is the abstract:

This paper answers fundamental questions that have preoccupied modern economic thought since the 18th century. What is the aggregate real rate of return in the economy? Is it higher than the growth rate of the economy and, if so, by how much? Is there a tendency for returns to fall in the long-run?Which particular assets have the highest long-run returns? We answer these questions on the basis of a new and comprehensive dataset for all major asset classes, including—for the first time—total returns to the largest, but oft ignored, component of household wealth, housing. The annual data on total returns for equity, housing, bonds, and bills cover 16 advanced economies from 1870 to 2015, and our new evidence reveals many new insights and puzzles.

Here is what I learned from the paper itself:

1. Risky assets such as equities and residential real estate average about 7% gains per year in real terms.  Housing outperformed equity before WWII, vice versa after WWII.  In any case it is a puzzle that housing returns are less volatile but about at the same level as equity returns over a broader time span.

2. Equity and housing gains have a relatively low covariance.  Buy both!

3. Equity returns across countries have become increasingly correlated, housing returns not.

4. The return on real safe assets is much more volatile than you might think.

5. The equity premium is volatile too.

6. The authors find support for Piketty’s r > g, except near periods of war.  Furthermore, the gap between r and g does not seem to be correlated with the growth rate of the economy.

I found this to be one of the best and most interesting papers of the year.


Thanks for pointing to this paper, will have to read and see how it compares to Triumph of the optimists and implications on cost of capital debates.

If you are predicting long-term equity returns of 7%, you are also predicting a large correction (downward) in stock prices. Few know this.

How far from 7% do you need to move to not predict a correction any more? 7.5%? 8.5%? I don't know, and I am genuinely curious.

@Ezequiel - according to no-load pioneer Bodie of Vanguard, the future rate of return for stocks, to avoid a correction, is around 3% he said a few years ago. So your money doubles every 25 years, not every 10 years as with 7%/yr.

I meant Bogle of Vanguard. I was thinking of the finance professor Bodie when I typed this.

He said real 3%, not nominal 3% IIRC. Anyway I have done the same analysis and I get real 3.5% or so.

@Larry Siegel - thanks, noted. Anyway inflation is almost zero for decades to come to nominal = real for the most part.

You have this backwards. With no correction, long term returns will have to be *lower* than 7%.

In the long run, the rate of return is dominated by the dividend yield (or, more accurately the payout yield, adjusting for stock buybacks). When stock prices are high, the yield is low. Right now you have to pay a lot for the stream of payouts. If stock prices fall, you can buy the stream more cheaply and long-run returns rise.

Here's a worked through numerical example that illustrate this idea:

"residential real estate average about 7% gains per year in real terms"


Residential real estate is in a constant state of decay, averaging probably 3% per year. Sum the component assets with useful lives of 5,7,10,30,50,100 years and a weighted average of 30 years is probably reasonable.

Technology exists to extend the lives of many components, but some go out of fashion resulting in their replacement no matter remaining useful life. Eg wall and floor covering.

Only select housing sees demand asset price inflation in excess of the constant investment required by housing. NYC vs Detroit, Cleveland, and thousands of Midwest towns, and mining boom towns.

I grew up in some Midwest towns where houses today have close to the nominal price half a century ago, and that's after significant investment over the decades to offset depreciation, aka decay. Two roofs, five water heaters, a furnace, a couple of paint jobs, new flooring, etc.

I considered returning at times, but the services that were robust half a century ago are below standards for then, and very low quality for today.

My guess is 80% of real estate in existence half a century ago is gone or undesirable or desirable because its cheaper than ever in real terms.

The value of "house" in Los Angeles is not the value of the house, but the value of the land on which one is allowed to build a house.

That is not true even in Los Angeles, and it varies greatly by location. In some places, nearly all of the value is in the structure.

Here is an article considering these issues, which concludes that 1/3 of real estate value in the United States is land:

I think mulp has an excellent point here.

all goes to show that the fairest real estate to measure would be the value of the land, not the value of the property.

anyway, if you subtract decay then you should also add rent

Return on real estate is not just added value of the house.
But also the annual rent earned. (Or - if the owner lives in the house himself - rent he would have had to pay if he didn't own a house).
Value added plus rent earned (and minus maintenance cost) might actually reach 7%. Though I agree: the figure seems somewhat high.

Seems like 2-3% return (price gains) based on inflation plus 4-5% cap rate (return based on rental return) is pretty realistic.

Real returns on structures of 7 pct is not possible. Real returns on land of 7 pct in very limited numbers of supply restricted markets might be possible. Is it possible the real estate data doesn't look at the negative carrying costs correctly.....? Estimating volatility of non mark to market assets is not that easy either.....

The high returns to housing could be explained by the very high transaction costs.

This is why regulatory obstacles preventing middle and lower-middle households from getting mortgages are damaging. Returns in low tier markets are higher than in high tier markets (price/rent ratios are lower). It's a great way for households of lesser means to earn high returns with low risk.

@Kevin Erdmann - "This is why regulatory obstacles preventing middle and lower-middle households from getting mortgages are damaging" - always the contrarian!? So I guess Fannie and Freddie were good things? Sorry I did not read your 100 part online narrative of the housing bubble, maybe I should.

I wonder if "home improvement" also creates phantom returns in residential real estate time series data....above and beyond maintenance. Especially post WWII.

Looking at longer horizons, the real returns on real estate are much lower, possibly as low as 0.5%, per Eichholz (1996) which looks back to the 17th century.

Of course real estate appears less volatile; it's less liquid. If you could mark a house's value by the nanosecond I imagine you'd see more vol in a residential house price index. This (and the associated tail risk) is exactly what fooled people into thinking that fixed income arb hedge funds have low vol.

The liquidity comment is on point. There are substantially more transaction costs both explicit and implicit.

Transactions are less frequent because of that and because of a host of other factors -- would you like to move every week? Would you like to shop for a new apartment building to invest in every month? Can I easily buy 0.01% of a big apartment building? And if your answer is "REITs", then perhaps that's part of the reason that REITs act more like any other stock than they act like individual pieces of real estate.

Or, more simply, is there anyone day-trading houses?

What is the definition of return on real estate though? A home is a consumption good. The only investment return that makes sense is the rental value return over costs plus appreciation, or the delta from home ownership cost to equivalent rental expense. Just looking at a price index is the wrong metric.

Why ignore the rate of return on productive capital (plant and equipment)? Isn't that the most important rate of return? I suppose not for the investor class, but it is for the working class. I know, one can't go to a source and find data on the rate of return on productive capital, which is itself inexplicable. The few economists who have studied the subject inform us that the rate of return on productive capital has been depressed for decades, which at least provides a rationale for the preference for alternatives such as financial assets (including cryptocurrencies!) and real estate (the latter becoming more like the former). Supporters of the Republican tax bill claim it will unleash billions in investments. Investments in what? Productive capital? Or speculative financial assets? I know, this is my hobbyhorse. And I know that stability of the world economy is maintained by central banks by manipulating the return on financial assets. Tim Wu, in an op/ed in today's NYT, explains the confidence in Bitcoin as more a reflection of the lack of trust in central banks than confidence in Bitcoin as a store of value. In another op/ed in today's NYT, Isaac Martin explains that the political playbook for tax cuts for the wealthy was written by Andrew Mellon, the Secretary of the Treasury in the 1920s: "Mellon squared the circle by inventing a supply-side argument: Cutting income tax rates would actually increase tax revenues. In particular, he said, cutting the top income tax rates would encourage rich people to pull their money out of tax shelters and invest in creating jobs." Mellon got his tax cut and the Great Depression soon followed.

Rayward: "And I know that stability of the world economy is maintained by central banks by manipulating the return on financial assets" - you are being fooled by randomness my friend. Money is largely neutral says the evidence (Fisher Black, Bernanke's FAVAR paper, etc). If only central banks had that much power. You sound a bit like those earnest but misguided Greeks I've met who claim the US CIA controls all the governments all over the world, and keeps Greece backward. If only the US had this sort of power I tell them, but they won't hear it.

Are you dull-witted. The point was that central banks don't maintain stability by manipulating the return on financial assets; indeed, they don't maintain stability! Ray, you are spending too much time in the jungle. You don't even know where the Spanish American War was fought. Hint: it's where you reside, or claim to reside.

@rayward - you were not clear counselor, look at the four corners of the instrument, and the passage I quoted verbatim. If you were being ironic it was not understood. As for the Spanish/Mexican typo I made, it was just that, and it's impossible to correct in Wordpress. This site, like Sumner's, should switch to a Simple Machines type forum where you can have a proper debate, in threads that are collapsible. And your trolling is noted, wayward!

Bonus trivia: Manila was founded by an aide to F. Magellan, who has an interesting story in Wikipedia.

We discussed this work on a personal finance website months ago. What has it cropped up here, now?

Isn't it obvious my German fiend? Because TC just read it. If a tree falls in a forest with nobody around to hear it...

Perhaps the aggregate value of owner-occupied real estate in the US has the same mean but less volatility than the aggregate value of US equities.

But with equities, a high degree of diversification in the asset class can be done easily and cheaply. By contrast, people typically own the house they live in and no other. That is the maximum degree of nondiversification. That makes home-ownership highly risk and volatile at the individual level, whatever the results for the national aggregate.

I will go out on a limb and claim that 1,2,3 are wrong due to methodological or data problems and therefore 6 is wrong and also the new "puzzles" are self-assembled.

I'd go out on the similar limb. But I can't pinpoint any obvious errors in the paper's methodology, other than the computation of comparing stock and real estate trading costs.

Two thoughts on housing returns seeming less volatile than they should be: 1. Housing isn't very liquid so it doesn't get repriced often and sellers will wait to sell if they don't think they're getting a good price. This is related to why private equity returns seems less volatile than they should. 2. Houses requires lots of maintenance. The rate of return will be lowered if this is accurately accounted for.

I didn't read the article, so they may have addressed these.

Would be helpful if Prof Cowen added a because to his assertions -- rather than just linking to lots of things on the internet.

So "one of the best and most interesting papers" because.. . .

Especially here where the paper is gated, the subject matter has been heavily studied and e.g. the long run return to equities and cross country correlations are well known.

Great paper, but I think it chronically suffers from success bias. The countries the paper uses are the countries that became developed, rather than those with sizable housing or equity markets in 1870.

The two largest economies [1] in 1870 were China and India - how did investing in housing do there? Similarly, in 1950, Argentina, Soviet Union, Brazil, Mexico and Indonesia and are in the largest 15 economies - I would imagine all asset classes in those countries performed very badly.

A more appropriate result is "If a countries economic and political system survives, then returns to housing and equity are both high"


There is a good point here, but it's not simply a success bias - I don't think China and India in 1870 were indistinguishable from the Western countries the authors picked. What I think is still interesting is the relationship between the asset classes - their claims about property v stocks are striking, if true.

You could write a different paper that also included non Western countries, and you would learn some more stuff, but surely it's OK to have some limits to your research ...

Yes, agree that limits are helpful. And that there are still useful results from the paper.

But I _do_ think this calls into doubt the r > g result. Economies which have political or economic upheaval generally have far worse asset returns than growth performance (e.g. if the state nationalizes all assets).

@Maximilian Roos-- I think the data shows in fact countries with low economic growth rates have high r nevertheless, so indeed r > g. South Africa comes to mind, as does the Philippines (for those lucky families with capital to invest). What you are referring to is the 'survivor-ship bias'. Obviously Czarist bonds preformed badly once the USSR took over.

If they didn't use a global market weighted composition than their results are kind of useless. I have seen much lower estimated real returns on equities when properly considering market weight - 4% for the US and 3% for the international market.

I take 3% real returns as a neutral to positive forecast for person equity returns. There is an irony that if the stock returns improve from their current market forecast of around 3%, it surely requires some P/E adjustment that on net has a likely negative effect on my future average portfolio returns.

I think 7% cited is not inflation adjusted returns? If so, 7% for realestate is reasonable, since it generally appreciates at inflation rate, ie 2%. But you also get 5% cap rate on rental income (rent - taxes and repairs), so 7%. Most residential owners put 20% or less down, so you get a 5x leverage on the 2% appreciation. So real returns are actually higher for the average owner.

There is a huge difference in rate of return from equities versus that from real estate and it can be ascribed to one word, "LOCATION." My humble 1955 split level house inside the Washington beltway is worthless as a house; it is only the property underneath it that is of any value. At this point, the physical home is a tear down for whoever purchase it. I receive 2-3 cards/letters a month from builders wanting my property so they can put up a $1.5M home on the lot (twice the current assessed value). There is zero reason for me to sell my home through a realtor when I can go direct to the builder and save on all the ancillary costs.

The same can probably be said for many other homes in other areas where there are high resale values. I've seen very lovely homes in formerly 'well to do' cities (Rochester and Buffalo NY are excellent examples) where real estate values are depressed or if they are increasing it is at a relatively low level.

Not having read the paper (I don't know if I will pay the $5 to purchase it as I'm not in the privileged class to get free NBER papers) is the question of whether the authors are just looking at real estate prices and have ignored the mortgage interest and cost of maintenance to come up with the 7% number. Take the example of a home purchased for $150K thirty years ago that is now worth $750K. If the owner had a standard 30 year mortgage and payed it down over the full 30 years at prevailing interest rates he/she would have spent another $150K in interest payments. Add into this other maintenance and perhaps upgrades such as a new kitchen and you have probably another $100K over the same 30 years. You now have spent $400K on a home that has not even doubled in price. The 7% return ought to have generated a much greater sales price. The only way this works is in rapidly increasing prices and selling the house after only a few years. I'm not even an economist but can see the fallacy of this kind of thinking.

Actually, 30 years ago, the interest rate was ~10%, and without refinancing, the payments would go: $30K down, $120K principal, $259K interest payments. Of course, refinancing would have decreased the overall interest.

Still, I think the calculation is the following:

Investment ($30K) for 20% down is the opportunity cost in potential stock investment. The principal and interest payments are a rent payment substitute.

In reality, this $30K has turned into $750K, and for the last half of the (refinanced) loan, the effective house payments were just for property taxes plus some change.

In 1987, Dow Jones was about 2000, today say 24K. 12x for buy and hold shares, 25x for buying a house.

@Viking - Yes, I know well what the mortgage interest was as we were first time home buyers in 1985 and had an ARM until we refinanced. But it is not $750K in reality. You have to subtract the mortgage interest, maintenance costs, updates to the dwelling, and I forgot to mention property taxes and yearly homeowners insurance. Those are all costs against the $750K that you would not have if you purchased an S&P Index fund at the same time you bought the house. $30K compounded at 7% is about $230K after 30 years. Under the house scenario that I am familiar with (my own!!!), I have probably done a little better than that after subtracting all the various payments but not much maybe a half percent. I don't think 25x that you mention can be correct.

@Alan - very true, the gentrified but otherwise scary looking NW corridor of DC is another example of land >> house. The paper was not gated, at least for me in the Philippines, try a proxy server from outside the USA (like medicines that cost pennies here but hundreds of dollars in the USA, maybe the paper is gated only for US residents).

Personally, I found this paper not that much different, except for the data maybe, than many papers done in the past in finance. Just another data point. An excellent book that is well worth reading, you can find it in softcopy on Piratebay, that debunks many cherished myths (including the dividend discount model, etc) is Coleman – Applied Finance (2017), a real contrarian from an Australian academic who has practiced in industry (he even worked in DC for a while).

Housing is local, so while good in many areas, not generally everywhere.

"2. Equity and housing gains have a relatively low covariance. Buy both!"

Would love to know what that level is. I have no knowledge of their data, but looking at this St. Louis Fed blog, it looks to me like housing (Wilshire REIT Trust) and equities (Wilshire Total Market), they look pretty darn correlated to me.

I also suspect that most people that buy real estate are owning their home. Buying in a local market is going to show a lot more variance than US real estate as a whole, so I think this advice regarding diversification is difficult to parse. Most people holding significant assets are already wholly over leveraged in local real estate. I certainly couldn't tell you what an appropriate level of US real estate holdings I should have to hedge against my DC-area housing wealth (and other equity and bond assets).

On page 39:

" the U.S., local (ZIP5) housing return volatility is about twice as large as aggregate volatility, which would about equalize risk-adjusted returns to equity and housing if investors owned one undiversified house."

Thanks for reading this. That answers my questions. I suppose one could argue that the authors are arguing that maybe one should have international housing exposure. I'd be even less sure about how to do that or how much I should buy.

Outside of an international REIT concept, that's just a horrible idea though.

Does it adjust for government subsidies of real estate development and purchase?

Schiller data showed that there was no increase in the inflation adjusted price of US houses for nearly 100 years before the recent housing bubble. The rental value of a house is generally assumed to be 5% of the price and property taxes are between 1 and 3 percent depending on the state. I do not see how you can get a 7% return on houses in the US with theses numbers, however the return on the s&p has been over 6%.

Maybe 5 plus 1-3 = around 7? Seems that indeed stocks beat housing post WWII, but for many people who leverage their first house, this point is not understood well.

Tyler and others, are real estate returns becoming more correlated too?

Thank you!

I hardly consider my residence an asset. My first home, an 850 square feet two bedroom from 1984 to 2000 property tax cost rose on average by over 11% compounded per year. Interest rates were plummeting during that time, but not property taxes. I probably paid double if one includes all the interest I paid to the bank at 12% (I did pay it off early).

Land and buildings are only assets to the government they belong. We can only license the "ownership" as much as someone "owns" a psl (personal seat license) for seat tickets at a stadium. We increase property's value so the governing body gets to charge ever higher rents into perpetuity.

A simple measure of this is the ratio of the property tax rate to the sum of the real interest rate and the property tax rate.

Suppose that house price at time t is P(t), the present value of future taxes is V(t), property tax rate is c and real interest rate applicable to housing property tax claim is r. V(t) = c*P(t)/r, or V(t)/P(t) = c/r. That is, the present value of taxes as a fraction of the combined market price of the house and the present value of taxes is V/(V+P) = c/(c+r). When the property tax rate is 1% and the real interest rate is 1%, what you actually purchase, when you think you are purchasing a house, is half a house. The other half belongs to the government.

One way to view this is that, as the real interest rates declined from 5% to 1% and property tax rates stayed the same and increased, the government confiscated 1/3 of all residential housing stock by not reducing the property tax rates proportionally to the real interest rates.

"1. Risky assets such as equities and residential real estate average about 7% gains per year in real terms. Housing outperformed equity before WWII, vice versa after WWII. In any case it is a puzzle that housing returns are less volatile but about at the same level as equity returns over a broader time span."

It is politically harder to tax/steal peoples' homes than impersonal and fragmented ownership in distant factories.

r > g to compensate for political risk and the evidence shows that political risk wipes out every dynasty on a pretty short timescale anyway.

Comments for this post are closed