Policy making in an uncertain world

Uncertainty was a key premise in Keynesian theory – and it remains an important variable in macroeconomics today

By Nilss Olekalns

Keynesian and other views on uncertainty

John Maynard Keynes, founder of macroeconomics, made uncertainty a key theme of his General Theory of Employment, Interest and Money. Keynes saw uncertainty as a market failure, most apparent in what we now call coordination problems. These coordination problems were a key reason for what Keynes saw as a tendency for economies to become trapped in a low employment-low output equilibrium.

The problem as Keynes saw it was the inability of firms to coordinate their investment decisions. Individual firms who perceived a lack of aggregate demand for their product would, understandably, refrain from investment spending. If enough firms acted in this way, there would indeed be an insufficiency of aggregate demand, in this case caused by firms’ lack of investment spending. On the other hand, if firms could coordinate their investment spending – that is, collectively agree to increase their expenditure – the very existence of this spending might generate sufficient aggregate demand in the economy to make that investment profitable. However, uncertainty about whether other firms will increase their spending discourages each firm from going ahead with its own spending. This is a classic coordination problem; one that Keynes thought could be resolved through government management of the economy to place a floor beneath aggregate demand. In Keynesian economics, policy acts in a very pragmatic way to resolve uncertainty.

There are other views of uncertainty. Fischer Black, for example, viewed uncertainty as something that could be to firms’ advantage. The idea here is that uncertainty is associated with volatility, and in times of volatility one often finds the largest possible returns from investment. In this framework, uncertainty, perhaps paradoxically, may yield desirable macroeconomic outcomes.

Examining different aspects of uncertainty

Here, I want to discuss two themes in my research on very different aspects of uncertainty. The first, which is collaborative work with my colleagues at the University of Melbourne, Ólan Henry and Kalvinder Shields, looks at how we might model the creation of uncertainty. Our research also looks at how that uncertainty might then affect the dynamic interrelationships that exist between macroeconomic variables. In contrast, the second strand of my research involves the destruction of uncertainty, attempting to strip away some of the uncertainty confronting policy makers when they are reviewing the current (and past) state of the economy. I am currently engaged in this research with Kalvinder Shields and another colleague, Kevin Lee, from the University of Leicester.

Let us step back for a moment and think about why a policy maker might be interested both in the effects of uncertainty and the ways to alleviate that uncertainty.

All decision making, including that undertaken at the national level by policy makers, involves some initial processing of past and present information. Against the background of this information, decisions are made that are expected to have particular outcomes in the future. The future, however, is uncertain. As the Roman philosopher and playwright Seneca said, ‘We let go the present, which we have in our power, and look forward to that which depends upon chance, and so relinquish a certainty for an uncertainty.’

The ultimate outcome of such decisions depends on what economists call the ‘state of the world’, but this only emerges with clarity some time after the decision is taken. For example, a decision such as that recently undertaken in Australia by the Reserve Bank to tighten monetary policy, will have an effect on the Australian economy that cannot be predicted precisely unless we know how the state of the world will unfold. Will China continue to boom, for example? The answer to that question will go a long way to determining whether current monetary tightening will push the economy into recession. Yet, although China’s continued growth is probable, it is by no means certain.

Modelling the effect of uncertainty on the economy

To what extent does uncertainty affect the economy? To answer this question, at least in a formal sense, requires a modelling strategy, preferably with a framework that is sufficiently general to encompass the origins of the uncertainty and how, once created, the uncertainty affects the economy. Empirically, measuring uncertainty is straightforward, especially if one is working with a properly specified model. Uncertainty is simply that which is not explained by the model and hence could not have been predicted.

However, this begs a host of important questions, not the least of which is: how does one know if one is working with the correct model of the economy? Given the number of competing models, this is not a trivial question. One approach has been the practice of model averaging. For example, it might be reasonable to advocate a policy change if, when the change is introduced into a series of competing models, none of the models predicts undesirable consequences.

An alternative approach is to work with a single model that is as general and flexible as possible; a model that can encompass a variety of possible structures of the economy. Using a very general type of modelling approach known as vector auto-regressions, we have developed a framework that:

Exploits the dynamic relationships that bind economic variables together;
Models how uncertainty is created through the arrival of new information (news) that could not have been anticipated on the basis of current or past information;
Allows that uncertainty to impact on the future realisation of economic variables; and
Differentiates between the effects of good and bad news in terms of how much uncertainty is created (with bad news creating more uncertainty than good news).

Our approach is inherently non-linear; that is, it takes account of the state of the world. This is entirely consistent with the decision making process noted earlier, where what unfolds after a decision is made depends on the state of the world. Thus, in our non-linear model, the state of the world matters as to how the economy responds to a particular piece of news. News that arrives in a recession, for example, may have an entirely different effect on the economy than if the same news arrives when the economy is booming. By way of contrast, most empirical economic models are not like this; they are linear in nature and, in such a framework, the response to news is independent of the state of the world. As our model includes the state of the world, it is a more general representation of how the economy responds to the creation of uncertainty.

Interactions between the share market and economic growth

One application of these techniques has been to look at the interactions between the share market and economic growth. Does uncertainty about economic growth affect the share market? Does share market uncertainty affect economic growth? And does it matter when in the business cycle the uncertainty occurs? Our findings suggest that the consequences flowing from the occurrence of uncertainty depend a great deal on the state of the economy. In particular, the arrival of news about economic growth (and hence the creation of uncertainty) has a much larger impact on the share market just after the economy has peaked as well as just after the economy has troughed. These results reinforce the view that the state of the world matters for the dynamics of the economy. To predict how a particular economy might respond to uncertainty requires knowledge about the state of the world.

Uncertainty about the state of the world

This leads to the second theme in my research – the uncertainty that exists about the state of the world. Do policy makers actually know, with certainty, about current economic conditions? The answer to this question is a fairly emphatic ‘no’. What policy makers have is an imperfect snapshot of the recent past and little or no information about what is happening in the present. They have no choice but to operate in a climate of uncertainty about the state of the world.

This uncertainty has its source in the technology available to statistics bureaus when they gather information about the economy in real time; that is, data available at a particular point in time. There are two sources of uncertainty in this connection. First, the information is subject to a time lag. Thus, the information on Gross Domestic Product for a particular quarter is available only in the subsequent quarter. Secondly, the information is revised, sometimes well after the quarter to which it relates. Both sources of uncertainty create problems for policy makers who require information on the contemporaneous state of the economy. Indeed, without this information, it will be difficult to predict exactly how the economy might respond if, for example, macroeconomic policy were to be changed.

The research with my colleagues Kalvinder Shields and Kevin Lee is a response to this element of uncertainty. The research involves nowcasting; that is, constructing a model that explicitly incorporates information about the timing of data releases and the nature of revisions that the data subsequently undergoes. By incorporating this information into the model, it may be possible to infer information about the actual state of the economy from the first release of the data. In addition, we have supplemented this with other sources of real time data – specifically, information from financial markets as summarised in the spread between long and short term interest rates, and the views of professional forecasters concerning the current state of the economy. This data, available in real time, is not subsequently revised, and may well be informative about the contemporaneous state of the economy.

The practical usefulness of this approach can be illustrated by reference to current developments in the US economy. At present, there is considerable public debate about the prospects for the US economy in the wake of the sub-prime mortgage crisis. Some commentators have expressed the view that the US economy is already in recession. The most recent data we have relates to Gross Domestic Product (GDP) in the December Quarter of 2007. This is first release data – we know that it will be revised. By adopting the modelling strategy as described above and then using simulation techniques, we are able to build up a picture of how likely it is that the preliminary data is consistent with the view that the US economy is currently in a recession. In this instance, we define a recession to be two successive quarters of negative GDP growth. This would be regarded as a severe slowdown.

We find that once we take into account (i) the dynamic interrelationships that exist between macroeconomic variables, (ii) the patterns that exist in the way variables are revised and the timing of their release, and (iii) the contemporaneous information contained in interest rate spreads and professional forecasts, the probability that the latest figures are consistent with a recession in the first quarter of 2008 is around 30 per cent. This is a large enough figure to give cause for concern.

Summary

Uncertainty is a pervasive feature in economics. I have described two modelling approaches that have, as their unifying feature, a requirement to take uncertainty seriously. Uncertainty matters for the economy and the way that macroeconomic variables evolve through time. Further, policy making requires techniques to peer through the uncertainty to establish the real facts behind it.

Top ^

A condensed version of his Inaugural Lecture delivered at the University of Melbourne on 11 March 2008. > Listen to Nilss Olekalns’ lecture

Professor Nilss Olekalns is Professor of Economics in the Department of Economics at the University of Melbourne.

Volume 3 APRIL 2008