I was a little surprised there's not more in land. "The only thing they're not making more of," etc. etc.
A little light poking showed that the most expensive land in the US is in high-end residential real estate, which is largely owned by rich individuals rather than by corporations.
I guess businesses know better than to build on expensive land. They build on cheap land -- or land containing something much better than dirt and rock that they can extract -- and then they make stuff on that land.
The only exception would be for hotels and high-end retail, I suppose? And it looks like hotels didn't make the cut and high-end retail was dwarfed by commodity retail.
So, uh, never mind.
Seems like public corporations make ownership decisions close to the finance-theoretical ideal where they minimize the assets they hold that aren't part of their production function to increase return on capital, and people who want to hold claims on rents buy them separately, consistent with the model I advanced in The Domestic Product.
Oh yeah, I was surprised by that too. I broke out land separate from buildings in most of my notes specifically for that reason.
Turns out the buildings (and, implicitly, building permits in high-density areas) are generally worth a lot more than just land itself. At least based on whatever black magic these accountants are using.
When economists talk about “capital assets”, they mean things like roads, buildings and machines. When I read through a company’s annual reports, lots of their assets are instead things like stocks and bonds, short-term debt, and other “financial” assets - i.e. claims on other people’s stuff. In theory, for every financial asset, there’s a financial liability somewhere. For every bond asset, there’s some payer for whom that bond is a liability. Across the economy, they all add up to zero. What’s left is the economists’ notion of capital, the nonfinancial assets: the roads, buildings, machines and so forth.
Can't stocks be worth a lot due to the profitable positive interaction between different things the company owns and rents, rather than due to their individual value? I'd think companies like Microsoft are to a substantial degree valuable because they've hired employees who've learned to collaborate to manage the technology sold by Microsoft.
Good question! Habryka also objected (offline) to the same block but making a different point: sometimes money just kind of "goes in circles" in the derivatives market, trading claims for claims for claims, and maybe doesn't actually end up in any nonfinancial assets.
IIUC the answer to both of these is double entry accounting. I am significantly less than 100% confident in my understanding here, but here's how I understand it.
Double entry accounting was designed largely for the specific purpose of making sure no money went "missing". To that end, a central constraint on all its conventions is that, for every debit, there must be an equal and opposite credit on the books. It's not showing "current market value of holdings" so much as "how much was spent to get the stuff". The "add up to zero" property is a constraint of accounting conventions specifically designed to achieve that property. Again, that's if I understand correctly; I'm significantly less than 100% confident.
That may apply to bonds (am not familiar with that), but I don't think double entry accounting is used to decide the value of stocks?
Their corresponding entry would be "equity" IIUC. So if Alice has $100 of stock in company X, that means Alice paid $100 for that stock (which now may have higher or lower market value, but the book still shows $100 worth for now).
If Alice bought the stock from Bob, who himself paid $80 for it, then Bob would have one entry for the buy and one entry for the sell, at $80 and $100 respectively. Those $80 and $100 entries both have corresponding opposite-sign entries in somebody else' books.
If the stock was originally issued to Bob for $80 by company X... I don't know the details of how that's recorded, but at a high level company X has an entry for equity and IIUC they sold some of that equity (or perhaps diluted the equity of existing shareholders to free some up, and then sold the freed-up equity). So, again at a high level where I don't know the low-level details, company X traded some equity (which is an entry in their books) for $80.
Let's say Alice buys 100 shares of Microsoft stock for $100. Then Microsoft implements a new management style that makes them much more effective, doubling the stock price. For emphasis on the new price, Bob then buys 1 share of Microsoft stock for $2. Alice's shares are now worth $200, but the extra $100 doesn't seem to have come from someone's transactions. This $100 would conventionally be considered capital owned by Alice, but the actual substance of this capital is purely based on the new management style of Microsoft, rather than Microsoft's assets.
I have some big machine (a combine harvester). The machine is worth a lot more than the individual components (gears, screws and so on) that make it up. Similarly, Microsoft is worth more than all the office buildings, patents etc that make it up put together.
So, value is not just in the number of physical things, but in the arrangement of them. I suppose that, ideally, the price difference between a bunch of gears and screws and a combine harvester should be equal to the cost of paying someone to assemble those gears and screws into one. So the price difference between Microsoft the company and all its stuff should be equal to the cost of hiring a bunch of managers to turn a similar amount of stuff into another Microsoft. Put that way ignoring that "arrangement value" does seem a bit artificial.
Alice' shares have a $200 market price, but the accountants aren't using that price on the books yet (unless they make a separate move to "mark to market"). Accounting values are not market values. This is one of the major reasons why I have those cautionary notes about accounting conventions throughout the post.
"Land is a minority of capital" is reassuring that this is mostly a summary of accumulated productive tools rather than of rent claims on natural resources rendered valuable by the productive use others can make of them. But it's in some tension with Gianni La Cava's claim that the increase in capital's share of income is largely due to increases in home values.
Presumably the solution to this paradox is that land values are mostly privately held, while public corporations tend to hold other forms of 'real capital,' so that rentiers still largely hold real estate, as they did when the term was coined. It would be interesting to learn whether privately held corporations' holdings are more similar to those of public corporations or natural persons.
That matches my expectation. There was a notable tendency for real estate to be a larger and larger fraction as I moved down the list, so I strongly expect real estate to dominate for smaller companies, and it certainly dominates for individuals. Though that's "real estate" including both land and buildings; I would still guess that buildings are usually a larger component than land even for private individuals.
That last bit seems wrong to me bc the "good location" premium is so large, e.g. https://www.crackshackormansion.com/. Davis and Palumbo (2006) estimated land value as 50% of residential real estate value, up from 32% in 1984, and home prices in aggregate have continued to rise for the same reasons.
Before we dive in, a few disclaimers. First disclaimer: this methodology is obviously quite biased. It completely ignores capital typically owned by the government (e.g. roads, water infrastructure) or individuals (e.g. houses, cars), and it’s weighted toward types of capital which are concentrated in fewer companies rather than dispersed over many. That said, the upside is that we know what the biases are; the methodology is systematic enough that there probably aren’t many unknown unknowns still hiding.
This also ignores human capital, which I believe (because an LLM told me) is valued at more than twice the value of the physical capital, (though it sounds like that estimate is going to depend a lot on some choices about how to measure things).
2
1
1That sure does depend on some very dubious choices. Human capital is rarely valued at all in standard accounting, it's the sort of thing I've mostly only heard priced in academic papers.
1Isn't it not valued in standard accounting because it's not an asset that's owned by the company in question; its owned by employees of the company? The company is ~ leasing the human capital by paying a salary.
I admit that it makes the analysis more abstract, and therefore more suspect, but we can totally compare the financial ROI of say Union Pacific buying a marginal train engine to Joe Shmoe getting a master's degree in engineering.
Does it seem crazy that most of human wealth is actually in the form of knowledge and expertise instead of in "stuff"?[1]
- ^
To the extent that a master's degree's primary mechanism of action is just signaling conscientiousness, intelligence, and conformity, then it does seem pretty crazy. It can't be the case that most of our real wealth is signaling, for reasons related to why all the financial capital nets out to 0.
But I think that Master's in engineering is a combination of teaching skills and certification of mostly innate properties.
1Labour income is the single largest component of GDP for many countries. And if capital is what generates cash flows, would this not be evidence for a larger human than physical capital component? My prior is with Eli’s comment, and not trying to estimate human capital will distort the analysis imo.
When a new dollar goes into the capital markets, after being bundled and securitized and lent several times over, where does it end up? When society’s total savings increase, what capital assets do those savings end up invested in?
When economists talk about “capital assets”, they mean things like roads, buildings and machines. When I read through a company’s annual reports, lots of their assets are instead things like stocks and bonds, short-term debt, and other “financial” assets - i.e. claims on other people’s stuff. In theory, for every financial asset, there’s a financial liability somewhere. For every bond asset, there’s some payer for whom that bond is a liability. Across the economy, they all add up to zero. What’s left is the economists’ notion of capital, the nonfinancial assets: the roads, buildings, machines and so forth.
Very roughly speaking, when there’s a net increase in savings, that’s where it has to end up - in the nonfinancial assets.
I wanted to get a more tangible sense of what nonfinancial assets look like, of where my savings are going in the physical world. So, back in 2017 I pulled fundamentals data on ~2100 publicly-held US companies. I looked at all the categories of balance-sheet assets, and picked out the nonfinancial types - things like inventory, patents, plant, property and equipment - then sorted, and found that 100 companies accounted for about half of the total nonfinancial assets. I then downloaded annual reports from all 100 of those companies (actually 102, for reasons I frankly don’t remember), and read through their balance sheets and notes to get a fine-grained idea of exactly what those capital assets look like.
This post is a compressed summary of findings from that project.
Disclaimers
Before we dive in, a few disclaimers. First disclaimer: this methodology is obviously quite biased. It completely ignores capital typically owned by the government (e.g. roads, water infrastructure) or individuals (e.g. houses, cars), and it’s weighted toward types of capital which are concentrated in fewer companies rather than dispersed over many. That said, the upside is that we know what the biases are; the methodology is systematic enough that there probably aren’t many unknown unknowns still hiding.
Second disclaimer: quantifying nonfinancial capital assets is tricky, because they depreciate. A 10 year old machine is worth less than a brand new machine. Different kinds of capital assets also depreciate at different rates. That means all the numbers in this post are sensitive to the details of accounting conventions - and unfortunately they don’t always use the same conventions (though at least different companies in the same industry usually use similar conventions for similar assets). The numbers in this post are good rough numbers on which to hang intuition, but don’t put a lot of confidence in their precision.
Third disclaimer: I have more background knowledge about some of these industries than others. I studied all of them enough to (I think) at least get a basic understanding of the capital assets involved, but there still could be unknown unknowns leaving me totally confused and interpreting numbers completely wrong. Please leave a comment if you see a mistake or if you want to add detail about a particular industry you know well.
Fourth disclaimer: I pulled these numbers in 2017. That was a while ago. Things change. Some things change more than others. Oil is a very cyclical industry; the power grid is pretty stable. AI datacenters weren’t a thing at all in 2017.
Fifth disclaimer: I pulled US companies, so naturally this is all fairly US-focused.
With that out of the way, we’re on to the fun part. Where is the capital?
Overview (With Numbers!)
The nominal grand total of capital assets (which is what I will somewhat-confusingly call the “real stuff” I was looking for) across all 102 companies was around $6.3 trillion. You should not take this number very seriously! It is largely an artifact of accounting conventions, and adding things up somewhat differently could change it by a factor of 2. But it does give a sense of the scale here, suitable for Fermi estimates.
The breakdown across my own off-the-cuff ontology of sectors looks like this:
Next I’ll do a walkthrough of those sectors and what their capital assets look(ed) like.
Oil - 25%
A bit more than half of the capital assets in the oil industry are in the “upstream” subsector, which basically means oil wells. Every well requires drilling a very deep hole, lining it with pipe and concrete, and (for the very large majority of the wells included here) fracking so that oil can flow more easily through the surrounding rock. That’s expensive, and there’s a lot of these things.
A lot of different companies own tens of billions of dollars each of those wells, with names like Anadarko Petroleum, Chesapeake Energy, Encana Corp, or Murphy Oil Corp.
Then there’s the rest of the oil industry:
There are also a few relatively large companies which vertically integrate wells, infrastructure and downstream processing; Exxon and Chevron are the biggest in terms of capital assets (at least according to these numbers).
Power Grid - 16%
Power plants, power lines, substations, etc. In the US, many different companies handle the power grid in different regions, and they all have both fairly predictable revenue streams and fairly steady capital expenditures on maintenance. The industry might as well be straight out of a textbook.
15 power grid companies were big enough to make it into my sample, and they mostly look pretty similar, just operating in different regions. The one exception is Berkshire Hathaway, which has almost half its capital assets in an otherwise-normal-looking power company subsidiary, but does a whole bunch of other stuff too.
Consumer - 13%
This category bundles together companies like:
The capital assets are what you’d expect: lots of buildings, equipment, and inventory. For instance, here’s Lowes’ breakdown:
(Note: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the company’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)
Telecoms - 8%
Companies: AT&T, Verizon, CenturyLink, Comcast, Charter, Sprint, T-Mobile, and Dish.
Looking into these gave me an appreciation for how the technical considerations of different telecom tech influences the capital expenditure and business model. AT&T is the oldest on the list, and their assets include an absolutely massive amount of “central offices”, i.e. the type of buildings where phone operators once sat, now populated by electronics.
A mobile-focused company like Sprint or T-mobile, on the other hand, needs tons of cell towers, but does not have a bunch of central offices. Comcast mostly owns coax cable originally intended for broadcast but now used to communicate in both directions for internet service, which is a whole different situation from AT&T’s legacy copper wire or Dish’s satellites and stations. (And presumably in the years since I looked into it, everyone involved has acquired a lot more fiber optic line.)
These are very different companies, for reasons mostly boiling down to their different legacy technologies.
Railroads - 8%
The “transportation” category in the pie chart above is actually all railroads. There are four companies in my sample: Union Pacific, Norfolk Southern, CSX, and most of the second half of Berkshire Hathaway’s capital assets.
As a sample, here’s how CSX’s assets broke down:
(Note: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how a railroad’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)
Healthcare - 8%
Think pharma companies like Pfizer, and medtechs like Medtronic. As a sample, here’s Pfizer’s breakdown:
Compared to Medtronic’s breakdown:
(Note, as usual: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the companies’ assets are distributed, but don’t compare the absolute total to other numbers in this post.)
Tech - 6%
Now we’re getting to smaller categories, so I’ll rattle through them a little faster. This is e.g. Google, Amazon, Apple, Intel, etc. Sample breakdown for Google:
(Note: not adjusted, relative sense, blah blah hopefully you read one of the earlier warnings.)
Industrial - 5%
Think GE, Boeing, or Caterpillar (the company which makes excavators, backhoes, and other fun construction vehicles). At the time of this sample, GE was winding down their brief foray into owning and renting out aircraft, so their numbers took me by surprise. Anyway, here’s the breakdown for Caterpillar, as a sample:
(Note: … ok just go look at the previous notes if you want the warnings for these numbers.)
Mining - 3%
Three companies - two with “gold” right there in the name, one mainly in copper. Here’s the breakdown for Goldcorp, no points for guessing what they specialize in:
(Note: y’know, I should probably actually copy the warning occasionally, so: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the company’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)
Real Estate - 3%
Three real estate holding companies. One of them is Simon, which owns a bunch of malls. It’s exactly what you’d expect, at this point.
Automotive - 2%
Ford and General Motors. Here’s Ford:
(Note: vo chugga depreciation, ne degada forsooth, uncompare, caution.)
Logistics - 1%
Fedex and UPS. Here’s UPS’ breakdown:
(Note: go check the warnings in above sections. Preferably one which actually explains the warning.)
Miscellaneous
Finally, a fun section. Three companies defied categorization: Carnival, Disney, and Visa.
Carnival has about half of the global cruise market (including Carnival Cruise Line, but also a bunch of other brands). As you’d expect, the overwhelming majority of their capital assets are cruise ships.
Disney, you’ve heard of. Their parks are the big item.
Finally, Visa, the credit card company. I didn’t look into what was going on with their assets enough to know what was going on; the large majority was some kind of accounting artifact having to do with a big acquisition.
Learnings
My biggest single takeaway from this project is infrastructure. If you ask “where does the capital end up?”, the main answer is infrastructure. Power grid, oil wells and pipelines and refineries, telecom lines and equipment, trains and airplanes and trucks. Then buildings and manufacturing equipment.
I also learned a long tail of how lots of different pieces of the economy work and why; every category above could be a whole essay on how its industry works and what it looks like. Indeed, that was originally my intent, but it’s been 8 years and that project will clearly never be done (at least by me… I would love it if somebody else tackled it, I would 100% buy that book).
Lastly: if you want to see my raw notes on the assets of all the companies, they are here.