The size of the shadow banking system estimated according to power law
medium.com
medium.com
1. Fit a model (in the case Zipf's law) to a semi-related field (the size of non-financial companies).
2. Find that the model doesn't fit reality in the field you're interested in (financial companies).
3. Posit that your model is correct and reality is wrong, and invent over $30 trillion worth of invisible activity to account for the difference.
Alternatively... your model is just wrong?
The question is: how could one find evidence that justifies or refutes the idea that the scale of firms does in fact fit zipf and that measure is being obscured at the tail?
Also, I'd point out that the original post did not differentiate between financial and non financial companies in the way you do between points 1 and 2, so your characterization of what fields are related or "semi-related" is your own creation and not part of the original claim.
Well, I think to find that evidence, you would have to take a measurement of the amount of shadow banking from one of these firms. Objectively, you can't do that, since there is no agreed upon definition of what the term shadow banking means.
Cough cough… The correct word is not "reality", but "currently well established models".
This is less obvious than you make it out to be.
It's a world of difference
Evidence based on assets capped at a lower value than the power law implies probably implies either regulation is limiting asset size or otherwise distorting the market, OR the books are cooked and assets are considerably higher but accounting games are being played. Perhaps thats a good idea for future financial planning if many of the "hidden assets" are actually worthless unrealized mortgage loan losses. Or maybe they're just crooks.
Anyway its a pretty big leap to go from unusually low assets to therefore total aggregate transaction volume must be artificially low.
For those who can't tell the difference between assets and transaction volume, this is a slight simplification but the assets of the car lot at my car dealer are about 200 cars, varies a bit, but they're all of probably about the same order. Now the total transactional volume is totally different at each dealer, some sell one car per day and some sell ten cars per day. The total annual transaction volume better closely match the total annual car production. However the assets on hand merely equal a simple division problem and don't mean a whole heck of a lot. Its like comparing apples and oranges or OSes and DMBSes. They're kinda sorta related but not the same thing.
Is that something new Medium are automatically injecting? If so, I don't like it. I do like the idea of being pushed to more content _after_ I've finished reading, not _during_ the middle.
Is this “This Week’s Top 5 Posts” Medium’s new deal, or is it yours? This is like HuffPost mid-article attention-bait. I can’t stand it.
I’ve previously equivocated Medium with an intense focus on single article content — no distractions. It was very confusing to find this totally unrelated link block here. I thought the article had suddenly ended.
What sort of click-through statistics are they trying to drive here? They don't have pageviews to pump on Medium (do they?), so what's the play?
- I disable ALL nags. If there's a modal, slider, fixed bar, or fixed social element, I nuke it.
- If it moves, I nuke it.
- If anything is labled "viral", "tease", "nag", or "social" there's really good odds it's getting nuked. Really, stuff like that just makes it too easy, and I may add some wildcards to my default CSS.
- Unless it's a premier site (e.g., NY Times, the Guardian), I'll nuke the "Recommended", "Trending", "Most Viral", etc., columns.
- I'll strip any 3-column (or otherwise multi-column) format into a single column of body text, generally about 40-50 ems (about 850-1000px) wide. The LH column generally becomes a full-width header, the RH column a full-width footer. If they survive at all.
- Bump all fonts to 15pt (body text), 14pt (usually for overview pages), 10-12pt for most navigation elements.
- Backgrounds to to white or very close to it, text to black or very close. #fffff4 background, #222 foreground, is about as much as I'll push either.
I've noticed that the result is much more soothing reading experiences, with far fewer distractions. It literally affects my ability to focus and peace of mind. Had my stylesheet manager (Stylebot) blow up on me a week or so back, having to deal with The Web As Published ... just drove me nuts. Most site's design simply sucks.
I've got over 500 sites styled to my preferences (many are just very quick and light changes, a few more involved). The results are ... mostly pretty nice.
Among my favorites:
Buzzfeed: https://plus.google.com/photos/104092656004159577193/albums/...
Unbuzzed: https://plus.google.com/photos/104092656004159577193/albums/...
A 3-column to 1-column design:
3 cols: https://plus.google.com/photos/104092656004159577193/albums/...
1 col: https://plus.google.com/photos/104092656004159577193/albums/...
And this seems to kill the "Top 5 posts" bits:
blockquote.pullquote, blockquote.pullquote + p, blockquote.pullquote + p + p, blockquote.pullquote + p + p + p, blockquote.pullquote + p + p + p + p, blockquote.pullquote + p + p + p + p + p, blockquote.pullquote + p + p + p + p + p + hr { display: none; }
Some of my edits are pretty rough. Most aim to please one key customer: me. And I do virtually no compatibility testing for devices, browsers, or even display settings.
Still ... could be useful. If only because exporting and uploading the stylesheets would give me a backup.
What would you pay for this?
Look at Market Caps of companies (something that we have public signals about): https://en.wikipedia.org/wiki/List_of_corporations_by_market...
The difference between #1 and #10 is <2x, and #3-#10 all vary by just a few percent. Power Law distributions are great, but they tend to fail at the top end.
"we find that financial firms dominate the top tail of the firm distribution by asset size"
But the implication that this works at the top end is either that the companies are hugely mis-stating their on-balancesheet assets or that there are other companies, much bigger than FNMA, that have balancesheets with multiples of the GDP of the States.
Neither of these explanations are plausible even if you believe that they're all lying, thieving scumbags. The numbers are just too big.
However, there is another explanation. That on-balancesheet assets isn't a good proxy for size for big financial institutions. And all of a sudden the problem is solved because on-balancesheet assets isn't a good proxy for size for big financial institutions.
Apart from anything else many of them run huge assets and liabilities off-balancesheet. BTW that doesn't make them secret, they hit the accounts just the same, it just means that the proxy in the study won't work.
And for what it's worth, this is wholly orthogonal to the size of Shadow Banking, how much stuff is hidden in the Caymans etc etc. Fascinating and worrying subjects in themselves.
I had to read this a couple of times before I realised what was wrong, but apparently "Econophysics" really is a thing:
"Econophysics is an interdisciplinary research field, applying theories and methods originally developed by physicists in order to solve problems in economics, usually those including uncertainty or stochastic processes and nonlinear dynamics. Its application to the study of financial markets has also been termed statistical finance referring to its roots in statistical physics." [1]
This is even scarier than the apparent fact that econophysics is a real thing. I'd like to know the history of the Forbes Global 2000 to get an idea over time of how dominant financial companies have been. I have a hunch they have always had a part (JP Morgan Chase anyone?) but surely this is a new phenomenon to be so dominated by companies that don't actually produce much.
GE is 4th, not 44th. It seems to be a typo.
market value, profits, number of employees all sound like better metrics to get the "size" of a company. if by "size" you mean the influence it has over the world.
If you deposit £50 pounds in the bank the cash becomes an asset of the bank and they create a matching liability to you because you probably want your money back at some point in the future....
Deposits are liabilities to a bank.
I was in IT at a stock trading firm a long time ago and the limited free training they provided for accounting was probably the most interesting non-job related on the job training I ever attended.
Banks manage other people's money, meaning they ought to lead in liabilities.
Bank assets primarily are the loans they extend to customers, such as mortgages and car loans -- in theory anyway, since they paper it nowadays to avoid carrying part or all of the risks. They extend these loans by leveraging the money in your CDs and savings accounts. In a sense, they borrow against the money they owe you.
When you deposit $50 in a bank and the bank loans it out, the loan is an asset of the bank, but the $50 in your account is a liability. Obviously if you just look at the assets the number looks really big, but the bank's net equity is the difference between the value of the assets and what it owes depositors.
It is interesting to me that 5 of the top 10 and all of the top 3 are financial institutions. By comparison to 2012, Only 4 of the top 10 were financial companies and ExxonMobil led the list.
People underestimate just how regulated the banking system is, despite all the talk about more regulation.
I agree with you about underestimation, but I think that the surplus of the regulations that we're talking about here is more to keep the member circle small, therefore it is in the benefit of the incumbent banks. So there is: 1. strong regulation if you want to get in and 2. weak regulation (at least for some banking activities) for keeping you in check afterwards.
1. Walter Willinger , David Alderson , John C. Doyle. Mathematics and the Internet: A Source of Enormous Confusion and Great Potential. Notices of the AMS. Volume 56, Number 5. May 2009. http://www.ams.org/notices/200905/rtx090500586p.pdf
If you want to avoid the term "economics" because most people associate it with "made up financial bollocks based on applying completely inappropriate models from unrelated disciplines", then why include the "econo" part.
Also, you appear to be trying to be modelling a financial phenomenon by applying a completely inappropriate model from an unrelated discipline.
As a concrete example: the Higgs result depends both on machine learning in the form of decision trees to minimize the data, as well as monte carlo simulation to generate theoretical data for comparison to what is measured. Experimental particle physics is very much now an exercise in machine learning and probabilistic simulation.
What I'm having a hard time with is the idea that markets necessarily have "laws" that are "discoverable" in the same way (which is what I think you're saying). They might do, and clearly some smarter people than me believe that they do, but where is the justification for this?
@contingencies mentioned [1] a George Soros quote about classical economics being based on a false analogy with Newtonian physics and markets are just the behaviour of individuals. Observations of particles tell you the behaviour of the particles, but the behaviour is fully determined by underlying physical laws. I don't see how this is true of behaviour of actors in economics, except in the obvious respect that the bodies of traders and the hardware of trading systems are subject to the laws of physics. What am I missing here?
I also wonder if there is a danger of building automated trading systems that are based on ideas from econophysics, and using their resulting behaviour are proof that the ideas were correct in the first place.
Newtownian physics was replaced by relativity, which is being muddled by quantum physics.
Physics is not cut and dried, so why should economics be?
To use your own example, the market is made up of the actions of individual actors, who are each governed by underlying mental models/laws (such as a probability field around making a trade, etc.). The OP is trying to discover what those models/laws ARE.
Yes, we use these methods to confirm that reality matches some candidate law.
I personally don't think it's very meaningful to talk about physical laws existing in some independent platonic sense, but the debate over the principle of induction and the epistemology of science is a side story here IMHO.
Ultimately what we care about is that a model successfully predicts reality in the future. I think a useful conceptual division is: with some models, the internal structure of the model gives us understanding of the physical behavior. Most of what we call "laws of physics" are in this category. We can look at the algebraic expression and understand symmetries, constants, conserved quantities, etc. A second type of model is purely algorithmic and utilitarian: it gets the right answer, but the internal structure of the model doesn't provide any comprehensible information or understanding. Most ML algorithms fall in the latter camp.
What muddies the issue is that experimental physics tries to validate models of the first kind, but now generally requires methods of the second kind. We're forced into these methods because of the complexity of what we're studying (particle physics can involve thousands of Feynman diagrams), the noisiness of our measuring equipment, and the rareness of the events we're searching for.
When I say physics methods are very relevant to wall street, I'm talking about the latter methods more than any attempt to find fundamental theorems.
In the past 100 or so years economics has completely failed to define itself as a science. You're using the wrong tense.
Anyway, we're drifting off-topic - I was bitching mainly about the title and a little about the misuse of Zipf's law. If the article was written in reverse, finishing with "this closely matches a power curve from which we can estimate x" then I would only have had a problem with the title "econophysicists".