Economic models generally look at shocks to the mean of a variable, e.g. what if interest rates are higher or lower on average. But what about shocks to the variance? How do shocks that increase uncertainty affect the economy?:
Momentous modeling, The Economist: ...[E]conomists ... start with a model of the economy, administer a “shock” to it—a sudden rise in the oil price, say, or some technological advance, or a cut in tariffs—and work out what happens to output, prices, employment and so forth. ...
Almost all these exercises deal, in effect, with changes in a number's mean ... (in statisticians' language, the “first moment” of its probability distribution)... Economists have spent remarkably little time working out what will happen if the world becomes less certain... Questions of this sort (about the “second moment”) are certainly more difficult to answer ... but important nonetheless.
They are important partly for the simple reason that people's behaviour may change if the world suddenly becomes a less (or more) certain place. Moreover, points out Nick Bloom, a British economist at Stanford University, events that increase uncertainty ... are fairly frequent. The most obvious recent example of a big second-moment shock is the terrorist attacks of September 11th 2001; but there have been plenty of others in the past few decades, from the assassination of John Kennedy to the collapse of Long-Term Capital Management and Enron (see chart).
Economists have not shied away from second-moment shocks altogether. Indeed, in 1983 ... Ben Bernanke published a paper analysing how investment would be affected by uncertainty. Uncertainty, he argued, could increase the return to firms from waiting rather than making irreversible investments. Mr Bloom has revisited the subject, with the fuller treatment that modern modelling and computing tools allow. ...
The model predicts, as does Mr Bernanke's, that firms wait and see what happens. Because the future is less certain, the ... value of waiting ... increases. ...[T]he net effect is that investment and employment fall. Productivity also drops... As uncertainty returns to normal levels, investment, employment and productivity bounce back.
Mr Bloom thinks that this model fits the aftermath of the September 11th attacks reasonably well. Net employment growth in America fell sharply in the three months following the attacks ... but rebounded in the first quarter of 2002. Investment also dipped and recovered. The perception that uncertainty increased also shows up in central banks' minutes of the time. One central banker spoke of some households and businesses entering “a wait-and-see mode...They are putting capital-spending plans on hold.”
For policymakers, says Mr Bloom, it is important to tell second-moment shocks, which seem not to last long, from the first-moment variety, where the effects endure for longer. If central bankers mistake a temporary rise in uncertainty for a permanent shock, they may, say, cut interest rates by more than they need to, with inflationary results. To make matters harder, first- and second-moment shocks are likely to come at the same time...
Are uncertainty shocks always transitory? Almost. ... Looking back over the past century, Mr Bloom can point to only one such episode: the Great Depression. Between 1929 and 1932, average stockmarket volatility was 30% greater than after the 2001 attacks. A prolonged bout of uncertainty—compounded by bad policy ...[had] catastrophic effects on investment and jobs. The direct macroeconomic consequences of September 11th, for all the deadly terror of that day, were much briefer.
In case anyone is wondering (and realizing few likely are), what about ARCH (Autoregressive Conditional Heteroskedasticity) models, don't they also provide a model of the variance? Why aren't they mentioned? The ARCH framework does provide a model for the second moment describing the evolution of the variance through time, but the equation describing the evolution of the variance is deterministic. As far as I know, within this framework it is not possible to shock the variance, the exercise considered in the article (adding a random shock to the variance equation creates identification problems for ARCH models - if this has been solved or if someone knows a way around this problem, e.g. modeling a structural break in the parameters of the variance equation, please let me know - Update: Please see comments on Stochastic Volatility models.).
So the ARCH framework isn't generically informative about the question posed in the article - a shock to uncertainty. However, ARCH models are useful for modeling financial market variables and macro variables such as inflation where periods of high and low volatility tend to come in cycles.
One question about the current economy which has seen a sharp drop in volatility since the mid-1980s (this is known as the Great Moderation) is whether the low volatility we are seeing is cyclic, or a change in the underlying trend. If it's cyclic, and there are worries that it is, a return to higher levels of volatility could create economic difficulties. If lower volatility is here to stay, then the question of what caused the change is still open. Was it better policy such as inflation targeting, innovations in information technology, or some other factor?
Finally, in thinking about these issues, it's also useful to remember the distinction between risk and uncertainty. These shocks do not hit the second moments directly, though it may be convenient to model them in that fashion. The exercise above is to have a shock of some type hit the economy with the consequence that uncertainty is increased. This adds an additional channel for the economy to be affected over and above the usual ways the shock would affect the economy's performance.