Connectedness

From a member of the Physics Department here at the UO:

Notional vs net: complexity is our enemy, Information Processing: The credit default swap (CDS) market, where AIG played, had notional outstanding value of about $45 trillion at the end of 2007. Of course many of these contracts are partially canceling, so the the net value of contracts in the market is much smaller than the notional value.

Unfortunately, the network diagram (network of contracts) probably looks something like this:

Imagine removing -- due to insolvency, lack of counterparty confidence, lack of shareholder confidence, etc. -- one of the nodes in the middle of the graph with lots of connections. What does that do to the detailed cancelations that reduce the notional value of $45 trillion to something more manageable? Suddenly, perfectly healthy nodes in the system have uncanceled liabilities or unhedged positions to deal with, and the net value of contracts skyrockets. This is why some entities are too connected to fail, as opposed to too BIG to fail. Systemic risk is all about complexity.

If bonds are issued to finance government spending, and if they are treated as new wealth by the private sector, they will stimulate new spending. However, if taxes are also increased at the same time, then the assets (bonds issued to finance the debt) and the liabilities (taxes) cancel each other out exactly and the bonds will be neutral, i.e. they will not stimulate any new spending since no net wealth is created.

Realizing this, people then asked, what if you give the bonds to the present generation as you run a deficit, and save the taxes for the next generation, wouldn't bonds be net wealth in that case? One group gets the benefits, the other the costs. Robert Barro answered the question in the paper "Are Bonds Net Wealth." He pointed out that if parents care about their children, then they will adjust their bequests to account for these kinds of changes in intergenerational distribution of assets and liabilities. Under the right conditions, e.g. perfect capital markets, the present value of all future liabilities will exactly match the present vale of all assets and no net wealth is created.

A key element here, though, is the connectedness of generations. Not everyone has children, for example, and Barro's mechanism works by putting the utility of children as an argument in the parents utility function. In the 1980s, in response to Barro's paper, I remember seeing a seminar given that attempted to estimate intergenerational connectedness. I can't remember exactly what the paper found after all these years, but the main point is that measures of connectedness exist. [In answer to the question, are bonds net wealth?, many people who have examined the empirical work take an intermediate position and use 50% as a rule of thumb, i.e. that 50% of bonds are net wealth, the other 50% is offset through anticipated tax liabilities).

Since measures of connectedness exist, and I presume physicists also have such measures (of complexity), I'm wondering if financial market regulators should start developing measures along these lines. Can we measure the connectedness of financial institutions econometrically? If so, can we also follow along the lines of the Hirfandahl index for monopoly power and develop guidelines for when a firm is too interconnected with other firms, so interconnected that it's failure threatens the overall system? Couldn't we then "break-up" the firms the way we do monopolies, "disconnect" the firm until it's failure wouldn't be so devastating?

As pointed out above, size alone isn't the key feature, the degree of connectedness (complexity) is also important, and regulators - as far as I know - don't have good empirical tools for assessing this aspect of financial markets.

Comments

[anne]

Right, I drew a spider web for AIG's reach.

[JeffF]

Did the now largely defunct regulatory barriers between various types of financial companies have the effect of reducing connectedness?

It seems to me that they may have done so quite strongly.

[robertdfeinman]

Since I'm one of the ones who has been arguing that "too big to fail" implies too big in the absolute, I'll put in my 2 cents.

I would expect that the number of links is roughly proportional to the size of an entity, so controlling for size would also reduce interconnectedness.

One could posit a situation where an entity has many small connections, think a savings bank with many small depositors versus a wealth management entity with only a few one, but the total dollar value is the same. In this case one would also need to factor in the "strength" of the connections. While the overall gross economic impact might be the same if the assets disappeared in both cases, the social impact would not.

I think this was implicitly taken into consideration when the FDIC was set up. Small depositors (even if numerous) are protected up to a limit where most are covered. Those who have more assets can use some of their resources to protect themselves via private insurance or by diversifying or otherwise.

Similarly interactions between firms tend to be proportional to the size of the dominant firm. GM buys more steel than Acme bicycles and a failure by GM would have a bigger impact on its suppliers than Acme. This is one of the objections to Walmart, it has monopsony power over many of its suppliers. There are documented cases where a change in their purchasing polices have caused firms to fail.

By limiting the size of firms, one would also lessen the risk that everything would unravel if one of them failed. If studies showed that this was not true, then there could be rules created about the degree of dependence of one firm on another.

I've not heard any good arguments as to why firms need to get bigger and bigger. Saving the salaries of a few top executives isn't justification enough, especially since really big firms have to add additional layers of management to handle things anyway. How many divisions does GE have and how many division presidents, VP's etc? What benefit does their credit division get from being chained to the jet engine enterprise?

[salao85]

I just listened to a programme on BBC radio 4 (http://www.bbc.co.uk/iplayer/episode/b00dgjn8/b00dgjm5/) where Andrew Hilton - Director of the Centre for the Study of Financial Innovation calls for a similar idea - retail banks (and I guess by extension insurers) go back to basics and investment houses become less not more regulated but are not allowed to get so big as to bring the whole system down. If they do they are broken up. He seemed to think that they would some how not all follow the same strategy which he didn't expand on, but said only crackpots talk about narrow banking at the moment; but people will come round.

[Bruce Wilder]

Gov't debt only has value in the present because we have the expectation that future cashflow from taxes will go to the bond holder. So, there must be an expectation of future tax expenses (to finance the payments to bondholders), which exactly matches the expectation of income of bond payments to bondholders.

If the bonds finance investment that enhances the taxbase -- increases the potential of future taxpayers to pay taxes -- then there may be an increase in wealth, but the bonds are not themselves representative of that increased wealth.

And, of course, if you Reagan-like engineer both the largest tax increase and the largest tax cut, but they fall on different economic classes, then the bonds represent a vehicle for the transfer of wealth, and have considerable net value to those, who own them. That is the wealthy recipients of cuts in taxes on capital income do not have to prepare to pay the increased taxes of workers paying increased payroll taxes; they can concentrate their political focus on stealing the Social Security Trust Fund, when convenient.

I am not sure any kind social solidarity needs to be analyzed.

[ken melvin]

Systemic is the operative word, isn't it? Like aspens, each tree can't count all the roots; only the aggregate plant can count the roots as its own. Before, it was, perhaps, more limited to the interconnectedness of corporate boards, but now, the board membership, the assets, the liabilities, ... are share by the aggregate.

[dd]

Is this current and is updated to account for the governmental node that now "owns" a huge proportion of CDS via Bear, Freddie, Fannie and AIG?

[ken melvin]

Maybe - From yesterday:

http://economistsview.typepad.com/economistsview/2008/09/paradigms-of-pa.html#c131026848

ken melvin says...

Seems to me:

If ‘A’ holds assets that include $1million of ‘B’ paper and that $1million includes $0.5million of ‘A’ paper, the asset held by ‘A’ is only worth $0.5million.

Further, if the other $0.5million of the $1million of ‘B’ paper is premised on ‘C’ and ‘D’ paper and one-half of each of these was premised on ‘A’ paper assets, then the asset held by ‘A’ is only worth $0.25million.

Und so weite.

Posted by: ken melvin | Link to comment | September 17, 2008 at 08:52 AM

[dd]

Sorry, see that government must be inserted in AIG nodes; but what of Freddie, Fannie and Bear? Wouldn't an update show a convergence and netting that cancels out some positions but sorely leaves the leverage intact?

[dd]

A government underwritten derivatives market is now possible and the model is the '34 Act. I have hope.

[Alex Tolley]

MarkThoma: "As pointed out above, size alone isn't the key feature, the degree of connectedness (complexity) is also important, and regulators - as far as I know - don't have good empirical tools for assessing this aspect of financial markets."

There has been some very good theoretical work using boolean network topologies - Stuart Kauffman "The Origins of Order". A key finding is that when average connections per node > 2, the network becomes unstable. At around 2 connections per node, the system exhibits a limited cycle of states. Below 2 connections, the system exhibits very few states. Although this work was done using boolean states and Gaussian distribution of links per node, anecdotally I have observed that the average 2 links per node is very common in biological networks. It would probably be worth redoing this type of analysis (if it hasn't already been done) using power law distributions of node links and observing the system dynamics (Barabasi is a thought leader here).

[Alex Tolley]

Practical policy based on very simple models might be problematic, but they may be useful starting points.

[dd]

Dear Tim and Ben and maybe even Al:
Sorry for all the criticism. Fractals, chaos theory and four dimensional thinking is hard; but thanks for hanging in there despite the current crash. It really might work. I'm excited about the future again despite the Tabal guy and even Buffet. Sorry about being slow on the uptake.
Thanks for the model Mark. A picture is worth and all.

[Creating or Borrowing]

"If bonds are issued to finance government spending, and if they are treated as new wealth by the private sector, they will stimulate new spending."

The terminology is preventing people from understanding what is going on. Issuing bonds means the same thing as borrowing consumer products from China et al, which can then be consumed domestically. That is, issuing bonds creates no consumer products, it just borrows them from China.

[Oupoot]

What is too big to fail? Or too connected to fail? Should the US then not consider that Goldman Sachs, Citigroup, BoA, Berkshire Hathaway, etc are currently too big &/ interconnected to be allowed to continue because of the risk they pose in the event that they fail? Why not split them up now to avoid a situation where the whole company comes crashing down if just one of their most interconnected units gets infected with this "cancer" spreading throughout the fin system?

IMHO, what the US should do is the exact opposite of what they think they should do. Why not bring back capital controls to stop the negative private capital flight from the US? Malaysia did it quite successfully in 1997...

[Ronald Rutherford]

Dr. Thoma:

You might want to check out this report:
http://www.iiasa.ac.at/Publications/Documents/IR-01-009.pdf

Intergenerational distribution questions seems to be quite important in Japan {I know you must be aware of that.} but should be something that the US thinks about seriously also.

[Brian Powers]

Nice idea. I reallly like it. Very clever.

Too bad those making these decisions are only looking for quick, short term solutions. Not sure they would get this anyway.

Still, I love it.

[Real Person from the Real World]

Because of interconnectedness applies not just in the financial risk sector, but also among people implies one problem is everyone's problem. The Great Chain of Being links people, jobs and parts of the economy. We need to get our economic house in order, and perhaps limit some of the connections to the global economy into modules, that won't pull everything else down if something goes wrong. Our people in India are working in the middle of the night, and as far as I can see, they aren't doing a very good job. Maybe following the clock is not the best paradigm to get the best out of workers. We also need health care and more access to education in the US. How can we do that if we beggar our own people via wage arbitrage and usurious credit card rates?

[rick.davies]

There seem to be two contrary notions of risk defined in terms of network structure.

1. In your argument above, excessive connectedness is seen as a risk

2. But in the management of individual investment portfolios, diversity of investments (i.e. lots of connections) is classically seen as a way of mitigating investment risk.

Is the key factor that makes a difference the size of the investment fund that has the large number of connections with others?