He argues that financial markets face the same trade-off as that faced by an electricity grid - a larger grid has greater capacity to limit the probability of blackout at any level (by drawing in on the larger numbers of alternative supply channels) but faces greater risks of a system-wide failure due to local failures (despite the presence of "circuit breakers" to isolate trouble spots). In the absence of integration, the failures would have been geographically constrained and the risk of "contagion" would be minimal. Prof Stiglitz therefore claims that diversification (through integration) and contagion are "different sides of the same coin - greater financial integration (especially if not done carefully) increases the risk of adverse contagion in the event of a large negative shock".
His model finds that it is not possible to always minimize the downside risk of integration (contagion) while preserving as much of the upside potential (diversification), especially with the likelihood of a bankruptcy cascade or systemic risk spreading in the existing credit market architecture. In the circumstances, it may be appropriate to introduce "circuit breakers" like capital controls when the markets face the risk of contagion and amplification of systemic risks.
He argues that the real estate bubble in the US was in accordance with the Greenwald-Stiglitz Fundamental Inefficiency Theorem which showed that even with rational expectations, so long as risk markets are incomplete, the market equilibrium will be inefficient. It also claimed that the effect of a shock can be amplified and lead to a process of trend reinforcement. In light of all this, imposing restrictions on certain sets of interactions (relationships) under certain circumstances may be welfare enhancing. He writes,
"Trade liberalization between two countries with negatively correlated outputs may reduce price volatility but increase income volatility, so much so that all groups in both countries are worse off. In an overlapping generations model, capital market liberalization impairs the extent to which a productivity shock at one time is 'shared' with future generations (as increased incomes raise savings and thus future wages) and thus can lower ex ante expected utility."
He rejects the assumption that the technologies/processes of global economic activity are convex and have concave utility functions (and therefore risk sharing or dispersal is always beneficial) as not correct. Information structures, learning processes, R&D, bankruptcy, and externalities, all give rise to a natural set of non-convexities. In the presence of these non-convexities, risk sharing can lower expected utility.
Therefore a firm experiencing a negative shock—forcing it closer to the brink of bankruptcy will have to pay higher interest rates, implying an increased likelihood of a further decline in net worth. Similarly, bankruptcy of one firm affects the likelihood of the bankruptcy of those to whom it owes money, its suppliers and those who might depend upon it for supplies; and so actions affecting its likelihood of bankruptcy have adverse effects on others - a bankruptcy cascade is set in motion. Externalities generated as a result of information costs and imperfections also contribute towards bankruptcy cascades.
His summary is very appropriate,
"In large non-linear systems with complex interactions, even small perturbations can have large consequences; even seemingly small changes in structure (introducing new 'connections' or contracts) can alter systemic stability. As our financial system became increasingly intertwined, through complex credit default swaps and other derivatives, too little thought was given to these matters, by the financial wizards that were creating the new products, by the bankers that were marketing them, by the economists that were touting their virtues, and by the regulators and policymakers who were responsible for ensuring the overall stability of the system."