|
The IMF is spoiling
to get even with its critics. Stung by claims that it missed
signals warning of the Asian crisis, gumshoe economists at the Fund
have been checking out the record of private forecasters. The
results are not flattering.
Bad forecasts have
been the butt of many jokes. "But the fact is," says Victor
Zarnowitz, "you can't do without them. So how good or how bad
forecasts are is a question of great importance." Zarnowitz is a
retired professor from the University of Chicago and now research
director at the Foundation for International Business and Economic
Research in New York.
New IMF research,
recently aired at a Federal Reserve seminar, should therefore cause
a stir. The paper, due out from the Fund this month, found that
private forecasts failed to predict a year beforehand any of the
recessions since 1989. There were 39, to be exact, in the countries
covered by the study. These setbacks included the US recession in
1991 as well as downturns in Hungary and Poland the same year, the
exchange rate mechanism crisis and related European recessions in
1992-93 together with the Mexican peso collapse in 1994-95 and, of
course, the Asian and Russian debacles.
This result came as
no surprise to at least one iconoclast. "Most economists are paid
to be cheerleaders," says A Gary Shilling, a New Jersey-based
forecaster well known for breaking ranks with the profession.
"Whistle blowers are unemployable."
Incentives matter.
Prakash Loungani, at the IMF, puts it this way: "If there were a
big reward from predicting recessions, more forecasters would try
to do it."
The IMF project
relied on data from Consensus Forecasts, a London-based
publication, and its offshoots: Latin American Consensus Forecasts;
Asia Pacific Consensus Forecasts; and, Eastern Europe Consensus
Forecasts.
Between 10 and 25
forecasters participate in the Consensus Forecasts for each major
country. But what are known as independent forecasters - those not
employed by banks, corporations, etc - are few and far between. And
many semi-government agencies - the Korean Development Bank, for
example - are part of Consensus groups for Asia. They have no
motive to stray from the party line. Sources familiar with the
situation told
Euromoneythat both the private Consensus as well as the IMF
tend to follow official government projections.
The Consensus groups
missed over half of the recessions in April of the same year when
they occurred. And about a quarter of the recessions remained
undetected as late as October. The track record for the IMF as well
as the World Bank and the OECD appears to be no better or no
worse.
That should cheer up
the folks at the IMF. News that private forecasters suffer from the
same blind spots could help them deflect charges that the Fund
can't make an honest, public forecast. Such forecasts are untenable
for the IMF, so the argument goes, because they would doom some of
the Fund's policy recommendations and programmes. The Fund, of
course, wasn't set up to serve as a forecasting service.
Not all forecasts are
equally off the mark. Private forecasts for developing countries
suffer from larger errors - relative to those for industrialized
countries - according to the study. Forecasts for developing
countries are the ones that tend to be too optimistic, over the
full business cycle. And forecasts for crisis countries are even
more off-target. Consensus Forecasts for industrialized countries,
meanwhile, show no "bias" one way or the other.
The IMF also
concluded that the consensus changed slowly for "behavioural
reasons". Forecasting, as JM Keynes pointed out, can be like
predicting the outcome of a beauty contest. The trick is not to
determine which woman is the prettiest, or even which one most of
the judges will think is the prettiest. Success, instead, comes
from being able to sense which girl most of the judges will think
that most of the other judges will say is the prettiest.
That's not entirely
unhealthy. "Forecasters get bolder," says Owen Lamont of the
University of Chicago, "as they get older and they become less
accurate as a consequence. The consensus is usually more accurate
than individuals." Similar patterns show up among mutual fund
managers and stock analysts.
Lamont points out
that independent forecasters are more likely to come up with jazzy
stories as a way to get attention and to drum up more business. Y2K
was a case in point. "It was idiotic to predict that Y2K would
cause a recession," he says. "Yet, some professional forecasters
did it. That demonstrated that the media like to talk to
forecasters who are exciting, not the forecasters who are
right."
Forecasters whose
income depends most on publicity will differ more from the
consensus. And Lamont criticizes perpetual pessimists such as
Shilling. "Even a broken clock," he says, "is right twice a day.
Shilling has predicted the last three recessions as well as 15
others that we didn't have."
Lamont, however,
thinks that Shilling and forecasters like him might still add
useful value. "It could be," he says, "that clients aren't looking
for accurate forecasts. If they're looking for a worst-case
scenario, then Shilling is providing something valuable." For his
part, Shilling concedes that there's often a fine line between
being a contrarian and taking an oddball position for the sake of
notoriety.
Still, GDP forecasts
can be all but irrelevant for many users. "I haven't found anyone
who can tell me what exchange the GDP trades on," says Shilling.
"The consensus may be correct most of the time, but it can also be
useless."
But Zarnowitz retorts
that short-term changes in variables with a more direct impact on
capital markets - like interest rates and inflation - can be even
tougher to predict. Other researchers have found that the US
Federal Reserve does a better job predicting inflation. But those
forecasts won't help investors. The Fed waits five years to release
them.
There are bright
spots, however. Monetary and credit aggregates serve as generally
reliable guides for predicting long-term inflation. And the
Consensus GDP forecasts have been reasonably accurate for large
economies, except at turning points in the business cycle. "Most of
the time," says Zarnowitz, "the economy expands and forecasters get
it right through simple methods like extrapolation of past trends
and so on, except at turning points when the largest errors
occur."
The consensus
therefore moves in a smooth rather than a "jerky" pattern. Lamont
believes that, overall, it does a little better than just
predicting 2.5% growth every year. But there hasn't recently been
any noticeable improvement in accuracy.
Zarnowitz considers
short-term GDP forecasts for the next few months or quarter to be
quite good much of the time. And the best of them are still all
right for about a year ahead. Peaks, as the IMF study found, are
not predicted well. "Troughs are predicted better," says Zarnowitz,
"even though leading indicators generally provide shorter
notice."
Zarnowitz argues that
people shouldn't blame forecasters for not predicting recessions
connected with such events as the 1990-91 war with Iraq. Those
jolts are all but unpredictable for the workaday economist, even if
they aren't primarily responsible for the business cycle. Another
vexing problem can be measurement errors by governments and other
data collectors. The resulting revisions have tripped up many a
forecaster.
Data problems have
been particularly severe in emerging markets. The Washington-based
Institute of International Finance recently reported that several
developing countries have upgraded information on foreign-exchange
reserves and debt. But that probably won't bring GDP forecasting
into line with standards elsewhere. Much of the bad forecasting for
those countries results from trouble predicting the success of
economic reforms. And that promises to remain a big problem as long
as the course of economic reforms remains uncertain in many
developing countries. The IIF, as a representative of major
international financial institutions, is primarily interested in
improving assessments of cross-border financial risk.
Forecast users should
take the following specific steps to avoid these pitfalls.
First, uncover the
forecasting cycle for each service. Most forecasters publish
monthly estimates. But they don't usually update their numbers
nearly that often. One way to improve on the consensus is to throw
out "stale" forecasts and then recalculate the average.
Second, look for
"skewness". That's the term quants would use to describe a
consensus forecast if the average is pulled toward one end of the
spectrum by a few extreme values. The sudden appearance of skewness
can signal a turning point in the business cycle.
Third, check for
discord among forecasters. The "standard deviation" for the
consensus represents one measure of uncertainty facing an economy.
It also rises near turning points.
"If you're really
concerned that we're at a turning point," advises Shilling, "you
ought to put a lot more weight on people who have some track record
in picking up turning points. More than anything, you're looking
for independence. It doesn't mean that they're going to be right,
but they've got a better chance of being right because at least
they've done it before."
Leading indicators
for the US and Germany, aren't predicting any decisive slowdown.
And forecasters generally expect Asia's crisis economies to
recover, although they seem somewhat uncertain. Only one bet seems
truly safe: if a recession strikes any of these countries, the IMF
and the Consensus won't predict it.
James Smalhout
|