Suppliers may go bankrupt because of the bankruptcy filings above. Suppliers may hold off on shipping parts without assurances they aren’t going to lose money in the bankruptcy. Either action idles assembly plants, cutting off the cash flow the car company needs to emerge from bankruptcy.

Yesterday, the Wall Street Journal wrote:

Troubled U.S. auto-parts suppliers were dealt a new blow Thursday when Chrysler LLC said it will temporarily idle most of its manufacturing during the bankruptcy process starting Monday.

Along with lost production, suppliers are at risk of having their payments from Chrysler disrupted as the auto maker’s finances are managed in bankruptcy court.

Two suppliers on Thursday refused to ship parts to Chrysler, forcing the auto maker to close a Warren, Mich., factory ahead of the planned shutdown, Vice Chairman Tom LaSorda said in a conference call with reporters.

Chrysler’s move threatens to push many suppliers closer to bankruptcy, and could ultimately lead to disruption in the flow of parts to healthier auto makers

I’m glad someone is paying attention to this danger.

Links: Chrysler Shutdown Could Push Parts Makers to the Brink – WSJ.com.

1. The 1965 blackout on the east coast of North America.
2. The 2003 blackout, again on the east coast.
3. A bank run is not by itself a cascade failure, but if it causes runs on other ones, it is.
4. The stock market crash of 2008.
5. The final fall of the World Trade Towers.

A cascade is a series of waterfalls that follow one another sequentially.

A cascade failure is the failure of a set of items that are linked, like the waterfall, into a sequence or progression. A good example is the game of dominoes. As Wikipedia says,  another common use of dominoes is standing them on edge in long lines then toppling the first tile, which falls on and topples the second, which topples the third, etc., resulting in all of the tiles falling. There’s also a picture of dominoes falling in that wikipedia article. Each one, as it falls/fails, transfers its weight to the next one, causing it to fail too.

Consider a set of dams on a waterway. A dam is constructed to pass a certain amount of water all the time and a greater (but still controlled) amount in times of high rainfall though a bypass mechanism called a spillway. The dams downstream are design to handle the flow from upstream spillways plus rainfall in their own catchment area.

If, however, a spillway is overwhelmed, the water may “overtop” the dam. Overtopping causes the flow to be faster and poorly controlled. This, by itself, may lead to failure of lower dams.

However, overtopping may also attack the dam structure and cause it to fail precipitously. Should that happen, the entire mass of water behind the failing dam will be added to the loading of the downstream dams, and the latent energy as well. Something very akin to a tidal wave will result. This is more likely to cause downstream dams to fail.

This, to me, is an essential part of a cascade failure. A linked system, already running at near-capacity, suffers initially a single point failure. The failing node transfers its load to the adjacent node(s). Not merely its static load but in many cases also it’s dynamic load. The domino doesn’t merely transfer it’s weight to the next domino, but also it’s falling momentum. If the weight of the falling domino isn’t enough to tip the adjacent one, the momentum of the fall will do the trick.

A cascade failure can be worse if the nodes are more connected. Rather than having merely one downstream link, one neighbour it can effect, like the waterfall or the dominoes, some networks have a sequence of 2 neighbours (the chains in a chain bridge, for instance) or more. Internet networks have, generally, many neighbours. A half-dozen or more is not uncommon. I will develop this connectedness idea further in another post.

Here are some examples of cascade failures taken from our everyday experience.