Can We Regulate Systemic Risk?

Systemic risks drive most crises, and pose a challenge for several reasons. First, they are not easy to detect with confidence, and are even more difficult to prove. Second, predicting the exact timing of a break point (when bubbles burst, markets lock up, and credit freezes) is, and will likely remain, beyond our ability. Finally, crises are highly non-linear events, which means that they occur without much warning.

Periodic outbreaks of instability impose high social costs on those who had the least to do with causing them. If repeated, this pattern may erode confidence in financial markets and regulators, which could well lead to heavy-handed regulation, the expansion of the state, and retrenchment from globalization.

But the problem is even more serious. The financial and economic crisis is morphing into a sovereign debt crisis in advanced countries. Financial and economic imbalance can lead to dangerous fiscal imbalance, as tax revenues plunge and social insurance and bailout expenditures rise. The International Monetary Fund suggests that as much as 75% of the “fiscal stimulus” in the advanced countries comprises non-discretionary counter-cyclical measures.

Undetected imbalances and systemic risk make fiscal policies seem prudent when they are not. Spain, for example, was not running a fiscal deficit coming into the crisis. But its revenues and expenditures were in part determined by a highly leveraged real-estate bubble.

Extreme fiscal imbalance can also lead to a growth trap in which fiscal consolidation has such a large negative effect on growth as to be self-defeating. Greece is probably a case in point. Eventually, the only way out is to reset the value of liabilities via restructuring or inflation.

If systemic risk can cause this kind of cascading sequential imbalance, then the “sovereign” needs to be alert, competent at identifying rising systemic risk, and able to take corrective action early.

We are about to get a comprehensive package of re-regulation focused on capital requirements and leverage, transparency, ratings and other sources of information, incentives, conflicts of interest and limits on the scope of financial firms, consumer protection, and resolution mechanisms. The hope is that such reforms will reduce the likelihood and severity of systemic risk.

But that doesn’t deal with global imbalances and other contributors to and signs of instability. In addition and as a complement to re-regulation, we need a comprehensive systemic risk monitor.

Some prominent policymakers and analysts, however, argue that oversight aimed at identifying and stemming systemic risk is futile. With incomplete models of risk dynamics and a complex and constantly changing global financial system, detection is, they argue, either impossible or so prone to error that the effort would be counter-productive. Asset bubbles are hard to identify with certainty.

If skeptics are right, then we should accept that we will periodically be out of financial and fiscal balance without knowing it in advance. Thus, we should also accept the need for much more conservative fiscal positioning than was thought necessary even three years ago.

In statistics and the theory of decision-making under uncertainty, errors are inevitable. There are two kinds. One error is to reject a true proposition; the other is to accept a false one. Let’s call them RWT (reject when true) and AWF (accept when false).

The issue can then be framed as follows: in the context of detecting and responding to systemic risk, which of the two types of errors has the higher expected costs?

Opponents of prudential oversight of systemic risk take two different positions.

One is that the AWF error won’t occur, because there are no reliable ex ante signs of rising potential instability. Looking for them is a waste of resources. Instability just strikes like lightning. This seems wrong. In the 2008 crisis, for example, some analysts issued warning signs, and some investors noted and responded to them. Admittedly, these warnings did not add up to an ironclad case, and they certainly didn’t predict the timing of the break. But the signs were there.

A second position accepts that there are warning signs, but that they are so unreliable that responding to them would do more harm than good. This implies that AWF has a higher expected cost than RWT, because there are a lot of false positives and/or the cure is worse than the disease. One can accept that there are costs to AWF. But RWT is costly, too, as we have seen, so the aggregate negative effect of the policy response to false positives (AWF) would have to be large to be convincing. I am not convinced.

There may be a deeper bias at work. In business and investing, choices under conditions of uncertainty are made all the time, and mistakes are routine. By contrast, developed country policymakers’ default stance seems to be that proactive or preemptive measures require a high degree of certainty, owing to a deep-seated belief that financial markets are stable and self-regulating.

If one believes that market instability is rare, then it is reasonable to refuse to act unless there is a compelling case to do so. In light of experience, the view that the financial system is only exceptionally unstable or on an unsustainable path seems at least questionable.

Based on new theory and experience, we may eventually conclude that policy responses to systemic risk are impossible to devise, and that the costs of the AWF errors are higher than the RWT errors. But we should at least conduct the experiment, assigning responsibility to a new or existing institution that has access to information, deep analytical talent in both financial and macroeconomic analysis, and is relatively free of conflicts of interest. The analysis should be made public and could influence perceptions of systemic risk and market behavior, thereby increasing self-regulatory capacity of the system.

Michael Spence, a Nobel laureate in economics and Professor Emeritus at Stanford University, chairs the Commission on Growth and Development.