Hacker News new | past | comments | ask | show | jobs | submit login
The physics of financial networks (arxiv.org)
133 points by Anon84 11 days ago | hide | past | favorite | 50 comments

from the top of the paper:

"As the total value of the global financial market has vastly outgrown the value of the real economy, financial institutions on this planet have created a web of interactions whose size and topology calls for a quantitative analysis by means of Complex Networks."

refer to Capital in the Twenty-First Century :

"The book's central thesis is that when the rate of return on capital (r) is greater than the rate of economic growth (g) over the long term, the result is concentration of wealth, and this unequal distribution of wealth causes social and economic instability."

regarding models, in a conclusion of Nature paper:

"Differently from other networks, here nodes have some ability to anticipate the future, including the future structure of the network and try to take advantage of it. Although in reality such ability is much more limited than what models in mainstream economics postulate, it represents a circularity in the model that is difficult to treat with analytical or statistical tools. "

so, participants use models and apply them to change behavior, making the system harder to model (!)

"so, participants use models and apply them to change behavior, making the system harder to model (!)"

I am ignorant here, but isn't this property true of iterated prisoner dilemma situations? Is game theory not an adequate model here?

That property is true of game theory problems, of which iterated prisoners dilemmas are a subset. And yes, game theory is the appropriate field to deal with those issues, however that does not necessarily mean that any solutions exist or are trivial for a specific problem.

Compare to a lot of engineering models being reduced to "just calculus"

I mean there are a lot of questions that can be posed, but are computationally intractable to answer. Think of the traveling salesman problem. Or worse - the halting problem.

refer to Capital in the Twenty-First Century

yikes, no thanks, keep it academic.

economics without politics is useless and pointless. economics is a social and behavioral science. the job of economists is to analyze and synthetize policies.

How's Picketty's work not academic ?

I have not read Picketty myself, but inferring .. Picketty is quantitatively substantiated, but also adds political descriptions and prescriptions; not academic in the sense of neutrality and reliance on accepted sources.

Honestly, calling for an academic treatment of economics like this, is less compelling for me even in retrospect, than quantitatively substantiated, current, analysis. How many people understand modern financial flows at scale? .. the topic of the original post.

I get your point now, however he does a really good job in Capital in the XXIe century to be purely descriptive and mostly rigorous (and why this work was regarded so highly even from rightist, neoclassical economists). His political philosophy contribution is mainly from his interviews and Capital and Ideology which is politically loaded because logically from his political "activism" he had to respond to the first book in which he wasn't. (as an idea, demonstrating descriptively how inequality works, how capitalism favors inequalities, people were expecting now what do you propose about it ? Which is then a ecopol book). I get where you came from though.

These guys' models can have multiple solutions. To me, that would indicate a problem with the model. But they are happy enough just to select the most optimistic solution.

I don't do financial modeling, but in all other areas of modeling it's definitely the case that there are many different "solutions" by which I'm guessing you mean local optima.

Having multiple solutions is somewhat equivalent to having different states of matter in physics. And economies definitely have very different modes, or states, that we call recessions or booms. Being able to have these sorts of local optima would seem to be necessary foe any model of an economy.

But I may not be understanding you, because we are using the same words in different ways...

This turns out to be a review paper by the same authors which only mentions the model I wanted to criticise. It's unlikely that one of the original papers will ever turn up on HN though! I was surprised to see this.

As I recall they began by adopting a structural model for debt valuation that assumes no dependence on any external assets other than the debtor's. Then they reintroduce full n-dimensional dependence through a different mechanism they made up to try and force the 1-dimensional solutions into a state of consistency. It's a bit of a mess.

Actually you're right. The presence of multiple solutions is not in itself enough to make me hate the model. The analogy relating different economic regimes to different phases of matter is potentially a good one.

I took a class of econophysics in grad school and one of the models with studied was a very simple generational model to explain inflation. It was originally discarded because it had an unrealistic solution with very high inflation, however it turned out to be a very good model for how hyperinflation can arise with a relatively small change in parameters. This was made particularly poignant given that the professor was Argentinian.

I have to read the paper thoroughly but I've come to the opinion that multiple solutions aren't in of themselves a problem if they are incorporated into inferences. A lot of Bayesian inference can be thought of as involving solution spaces, for instance, where you're assigning a probability to each solution.

Not disagreeing with you, just wondering out loud if it could be revisited with other methods.

As statistician George Box said, "All models are wrong, but some are useful"

> "As the total value of the global financial market outgrew the value of the real economy" - the first sentence

What the hell is this supposed to mean? What is "value" here, can financial markets conjure more houses, crops or datacenter capacity? Or does it mean that "market prices of securities" have outpaced the "market prices of" .. "the real thing"? Like, seriously, what?


The market prices of securities, and (critically) the returns on those securities, are now higher than the book value and returns of the cash invested in companies whose shares back those securities.

The market value of stocks and corporate bonds is pretty much always larger than the book value of the underlying companies. If this is not the case you've found yourself in a 1931-type of a bargain.

The delta between the book value and the securities pricing is the estimated future profits. Now how exactly did you arrive at the conclusion that the future profits estimation is too high?

I didn't come to that conclusion, and you are ignoring the part of my comment pointing out that the critical switch is that the _returns_ on securities now exceed _returns_ on actual businesses.

But, since we're now talking about overvalued securities:


You might be confusing Book Value of a business with it's value as an asset, and how returns are measured. "return on actual businesses" is impossible to measure with any precision.The time horizon chosen for returns on securities is usually short. So, the returns on securities will almost always seem too high or too low compared to what is happening in the underlying business.

The market returns are calculated over a fixed and usually short time period and are easy to calculate. Because the markets in question are expectations markets the prices move around a lot and hence the measured returns move around a lot.

Compare that with defining "returns on actual businesses" in a way which is both sensible (ie accurate) and has even a reasonable level of precision - it's impossible. For example, a commonly used and easy way to measure it is Return on Equity. Net Income / Book Value. Book Value is an accounting number based on the past. The "true" value put on an asset by investors is based on guesses about the future of the business and said business's ability to put cash in the pockets of the owners. Net Income misses important value changes - consider Intel, who have been making crazy decisions for years which allowed AMD back in the x86 server game. Intel don't have to report a "we did stupid things which will cost us billions in 10 years time" on their income statement for a given year. So, RoE isn't good and other similar numbers are similarly bad.

So, your statement about returns on securities should really say: "for the last few years, returns on securities appear to have exceeded the changes in value for the underlying businesses". To which a reasonable response is: "Yup, it happens sometimes".

You appear to be under the impression that there is a strong causal/correlative relationship between the value of a security and some measure / function of the value or potential value of the underlying business.

I would argue that there is an increasing amount of evidence that statement has gone from "plausibly true" to "pretty obviously not the case, for many securities."

The underlying problem, regardless of the terminology or metrics being used to estimate value, seems to be that those with $10b in cash seem increasingly inclined to use it to speculate using various financial instruments, rather than e.g. building a manufacturing plant or a new housing development or apartment building. The growing size of the financial sector relative to the rest of the economy is the evidence I would put forward that "the returns on securities are exceeding the returns on actual business."

I am under that impression - because most of the time most of the securities are priced in a reasonable range. There have always been a small number of weird situations where securities appear to have little relationship to the underlying. Right now there are a few - AMC and Gamestop for example. Sometimes, for short periods of time, there are many securities which are out of whack - 1999 for example. Go back 100 years and you'll see a number of periods where many securities prices were out of whack. Without hard evidence, but a decent amount of anecdata, common sense and experience I'd suggest the following: 90% of the time 90% of the securities are priced within 10% of "true value".

IMHO assuming that 'true value' even exists is very bold. you'd have to integrate all possible timelines into your valuation, modeled with very fine precision, and then you're lucky if your model converges to anything other than 0 or infinity...

Do you mean it's basically impossible to precisely calculate? If so, yes. But I believe (and many smart folks who have made a lot of money doing so also believe) it's possible to accurately give a range. Like, company X is worth between $900 million and $1.1 billion.

I mean that even such a range doesn’t really say anything because it doesn’t describe what is included in the valuation. Is it nav, npv or are some future cash flows also included? It’s up to the person giving the valuation to describe.

In practically every case when talking about a publicly listed company, such a number is referring to a fully diluted market capitalization based on the net present value of the cashflows which would be received by the common stock holders. The cash could be dividends, it could be buybacks, they could be a result of regular cash coming out of the business, or a sale of assets - it doesn't matter. There really is no confusion amongst people in the industry.If they are talking about something different, they will explicitly say so (the easy example is Enterprise Value, or EV).

How did you arrive at the conclusion that your estimates were correct in the first place? That argument cuts both ways.

If I lent you 1 trillion dollars and you lent it right back to me (because I don't actually have it), we'd both have 1tn in assets and debt. That nets off to zero.

But if you want to make up bullshit, you could pretend it doesn't and that you and I are now the world's largest economy, trading 2tn a day.

It's a great way to make your maths easier (but wrong) and your headlines much more attention grabbing.

Suppose you have some expectation about the closing value of the S&P500 in a month's time. You can bet on it using a variety of financial instruments, e.g. options and futures.

How much you bet is up to you and your counterparty; the amount is not in any way constrained by the actual value of the companies in the S&P500.

And that's how the size of the derivatives market can look like this when compared to the stock market:


You are right in that the amount isn't constrained. But the net value of a derivative contract is zero. Ie, it's not right to say the "value". The term "notional value" is used for this reason. To further make the point - there isn't always an asset underlying a derivative - for example there is a notional value of weather derivatives.

In summary, the word "value" gets thrown around loosely.

for good reason. it is impossible to define strictly in general.

There isn't a good reason to throw around the term loosely when talking about a specific field where the terms are quite well defined, like the one I gave.

Folks should use the right terms if they genuinely want to communicate ideas.

The stock market is $89T and the derivatives market is $11T.

Notional value is somewhat meaningless, tomorrow you could buy a single out of the money SPX 0DTE option for $100 that has a notional value of $423,900

market prices of securities don't really reflect the value of 'real things'. market prices reflect capital flows. sometimes the two align - but this also depends on the definition of value you use. (you frequently hear that somethings trades at X multiple of ABC and Y multiple of DEF... and those multiples are wildly different across the whole market, yet prices are stable - because capital flows balance out.)

A fancy way of saying an equities bubble.

I guess I have to explain. The reason it’s a definition of a bubble is due to compounding. If the economy is growing at 1 to 3% while equities are going up 7 to 10% a year they diverge slowly at first but the divergence is exponential. What ends up happening it takes far more debt to sustain this bubble. Interest have to keep doing down so the payments stay relatively the same but at some point rates can’t go lower or a shock causing the ability to pay goes away. Then the bubble pops causing a liquidity problem(2008, March 2020) causing a massive sale off of the most liquid assets. So far the governments of the world have been transferring the liabilities of the most toxic liabilities from banks, pension funds, etc to the Central banks. It is an asset swap and doesn’t inflate the money supply directly but what it does do is tell banks lend all you want if you blow yourself up we will bail you out. This further inflates the debt bubble since banks now believe the fed has backstopped their loans there by limiting downside risks. This money being created has to go some where so it goes into asset purchases. Then you have margin and loans based on the assets further inflating then bubble. This creates a feedback loop since the increase in the asset prices increases the amount of money available to borrow which is often used to buy other assets further increasing the price.

Yup, it's literally just "being bearish". If you want to see people "being bearish" or "being bullish" just flip on the CNBC or something and listen.

You either claim that the interest rate (or various risk premia) are too low, or that the estimates of future profits (for whatever reason, higher COGS, higher cost of labor, higher taxes, under-measured depreciation and higher capex, natural calamities, lower GDP and thus lower revenues, or really any other reason) are too high.

Not really. The market is not a zero sum game. Debt, loans, financial products can benefit everyone

It's a lazy, incorrect statement.

Was hoping to find out who the next LTCM or Bear Stearns is.

Not again. Econophysics tried hard 15/20 years ago but really made no substantial contribution to our understanding of markets.

Seems like a strange sentiment, evaluating a modeling approach based on historical analogies vs its own merits. If we're playing the historical analogies game, people were saying "not again" about Neural Networks up until mid/late 2000s.

Eric weinstein has tried to apply physics concepts to economics , such as gauge theory. I dunno if he succeeded. I remember it generated at the time a lot of hype

Ok, we've changed to arxiv.org from https://www.nature.com/articles/s42254-021-00322-5, which is hardwalled. Thanks!

preprint version [available here](https://arxiv.org/abs/2103.05623)

Ok, we've changed to arxiv.org from https://www.nature.com/articles/s42254-021-00322-5, which is hardwalled. Thanks!

Hard paywall.

Changed now. Thanks!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact