This article came out in the Deseret News on Tuesday, October 18th.
Last Monday the Nobel prize in
economics was awarded to Tom Sargent and Chris Sims. Both are well-known macroeconomists and both
have worked on economic issues relevant to the 2008 recession and recovery.
Forecasting the future of the economy
is tricky business. For one thing it is
very complicated, with millions of goods and services changing hands. Another reason is that it is subject to
changes in the economic environment that are not economic in nature; weather
and politics being two good examples.
Forecasting how the economy will behave requires simplifying models that
capture most of its features without adding too much complexity. Over the years, economists have developed
increasingly sophisticated ways of doing this.
By way of analogy consider the portion
of U.S. Highway 6 that runs between Spanish Fork and Price. I drive this stretch of road on occasion on
my way to the San Rafael Swell. The road
goes up Spanish Fork Canyon, over Soldier Summit, and down Price Canyon. It is necessarily winding and steep in many
places. Suppose you were tasked with
forecasting the fate of a convoy of vehicles traveling over this road.
A simple first stab at the problem
might involve using elementary physics.
The vehicles have given weights, they travel at certain speeds over
different portions of the road, the road's gradient and curvature are
known. Based on this information you
could, with some effort, derive a forecast for the progress of the convoy.
However, to improve your forecast, you
might also consider the weather. Unfortunately,
the weather is changeable. You have a
general idea of conditions, but the specifics at each point on the road are not
known. Furthermore, these conditions can
change unexpectedly. You need to make a
best guess and factor this into your forecast.
You will also need to update it as the convoy progresses and available information
changes. The same principle applies to
other factors like the mechanical condition of the vehicles, and the mental
condition of the drivers.
When you make your forecast you realize
that it is only a best guess. It is
subject to change due to factors that are difficult to predict.
If you had some control over the
highway or the vehicles you might be able to reduce the chances of a serious
slowdown or pileup. Suppose you had a
radio controller that could uniformly boost or reduce the amount of fuel all
the vehicles consume. If conditions
looked dangerous you could dial down consumption of gas, slow the convoy down,
and reduce the chances of something bad happening.
This type of forecasting and policy
recommendation corresponds roughly to state-of-the-art economic forecasting
prior to the introduction of Rational Expectations theory thirty to forty years
ago. Tom Sargent was an important
contributor to that literature. Chris
Sims' contribution was to develop statistical techniques that identify how
economic variables influence each other as time progresses.
When you cut back gas consumption you
assumed this would make the cars go slower.
However, drivers are not automatons and they adjust their driving
behavior based on the conditions they observe.
They do this by gathering all the relevant information they can: direct
observation of the road through the windows, listening to the radio, talking
with other drivers on cell phones, etc.
When things look dangerous, they slow down on their own. If you dial down the gasoline flow, drivers
will simply push harder on the accelerator to compensate and will maintain the
speed they think is best. . Ignoring the responses of the drivers
(decision makers) in the convoy (economy) to your manipulation of the gas flow
(economic policy) gives bad forecasts.
Good forecasts will incorporate these rational responses.
The 2008 financial crisis and recession
have been held up by some people as evidence that Rational Expectations is
incorrect. If decision makers are gathering
information and processing it effectively, how could they have missed the
subprime meltdown? Why did they ignore
the warnings of those who were warning against just such a meltdown at the
Go back to the convoy. Suppose you have a fellow forecaster who
believes the brakes on a large semi-truck are about to fail. As the convoy makes each turn along the route
he announces over the radio to all concerned that the semi is about to run out
of control, tip over, and cause a massive pileup. However, for the first several turns the
brakes hold and the convoy continues on unharmed. Eventually, everyone discounts his
predictions of doom. If the brakes
finally do fail, it comes as big surprise to almost everyone. Why did drivers not foresee the crash? Why did they ignore the voice on the
radio? Because the problem was small,
subtle, not readily apparent to anyone but an expert, and the exact moment of
failure was largely random.
The problem is not that the forecasting
methodology is wrong. Rather, it is that
unexpected or difficult to predict events can in some circumstances have huge
consequences. They are easy to see in
hindsight, but not so easy to see before they happen. One message the Nobel committee sent was that
rational expectations is still an important piece of economic theory.