Macroeconomic Forecasting: Methods and Pitfalls

1
Chapter 20

• Macroeconomic Forecasting Methods and Pitfalls

2
Useful Elements of Macroeconomic Forecasting for
Business Managers

• Unexpected events are, by definition, not
  predictable. Yet macroeconomics can still say
  something useful about what happens after these
  events. Even a totally unexpected shock is
  likely to set in motion a predictable series of
  events.
• To a certain extent, various leading indicators
  can indicate when trends are likely to shift or
  be reversed, although no indicator is 100
  reliable.

3
What Sorts of Factors Are Predictable?

• We have mentioned earlier that an inverted yield
  curve is always followed by a recession the
  following year.
• Sales of cars and housing will follow changes in
  short-term rates with an average lag of about two
  quarters.
• Changes in the stock market will affect capital
  spending with a lag of two to four quarters.
• Changes in the value of the dollar will affect
  net exports with a lag of three to six quarters.

4
What Sorts of Factors are Not Predictable?

• Obviously truly exogenous shocks, such as wars
  and terrorist attacks.
• Major fluctuations in energy prices used to have
  a significant impact on the economy, but as firms
  learned to hedge against these changes, they now
  have only a small effect.
• The impact of changes in income tax rates is
  mixed. It depends on the phase of the business
  cycle, whether the changes are seen as permanent
  or not, and what happens to government spending
  at the same time.
• Similarly, changes in corporate income tax rates,
  including investment incentives, have a mixed
  record.

5
Macroeconomic Forecasts with Econometric Models

• Not much used any more
• They didnt work for several reasons
• 1. Incorrect underlying theory
• 2. Instability of underlying relationships
• 3. Errors in econometric method
• 4. Inadequate and incorrect data
• 5. Tendency to cluster around the consensus
  forecast
• 6. Inability to predict exogenous events
• 7. Erroneous assumptions about policy variables

6
Are Econometric Model Forecasts Likely to Improve?

• Suppose these errors were fixed. Would the
  models work any better?
• Probably not.
• Monetary and fiscal policy have generally
  improved to the point where future recessions are
  likely to be caused by exogenous shocks, which by
  their very nature are unpredictable.
• If it appeared likely that a recession would
  occur the following year, the Fed would probably
  take steps to ward off the downturn.
• As a result, this chapter focuses on
  non-econometric methods of prediction

7
Non-econometric methods of forecasting

• Consensus forecasts
• Financial market surveys
• Leading indicators
• Surveys of consumer expectations
• Surveys of manufacturing activity
• Surveys of capital spending or inventory planning
  

8
Consensus Forecast

• Best known is Blue Chip Economic Indicators,
  other surveys give approximately the same
  results.
• For real GDP and the inflation rate, have about
  the same errors as a naïve model that says the
  change this period will be the same as the change
  last period.
• Has missed all the recessions.

9
Surveys of Upcoming Indicators

• They are often seen in the financial pages and
  screens these days.
• Surveys are designed to predict upcoming monthly
  and quarterly indicators.
• Obviously you cant make any money trading on
  this information.
• Chasing a will o the wisp. Preliminary data
  themselves are random.
• Dont waste your time.
• Even if some brilliant financial analysts
  actually found the key to predicting (say)
  employment or inflation, they certainly wouldnt
  tell the rest of the world about it.

10
Leading Indicators

• As previously mentioned, never miss a recession
  (unlike the consensus) but have predicted several
  downturns that never existed.
• Nonetheless, they should not be ignored. This
  index did a very credible job of predicting the
  2001 recession at a time when most economists
  missed it.

11
Leading Indicators, Slide 2

• Isnt it possible that not all components of this
  index are created equal, and some are better than
  others? In that case, why not follow the better
  components and ignore the poorer indicators?
• To answer this question, lets look at what is
  included in the leading indicators.

12
Components of Leading Indicators

• . 1. Length of workweek for production
  workers in manufacturing
• 2. Average weekly initial unemployment claims
• 3. New orders, consumer goods and materials
• 4. New orders, nondefense capital goods
• 5. Vendor performance (percentage of firms
  reporting longer delivery delays)
• 6. Building permits
• 7. SP 500 index of stock prices
• 8. Real M2 money supply
• 9. Yield spread between 10-year Treasury note
  yield and the Federal funds rate
• 10. Index of consumer expectations (University
  of Michigan)

13
Leading Indicator Components, Slide 2

• Some of these are outdated. The M2 money supply
  is not closely correlated with economic activity
  any more, and loans and credit are coincident
  indicators.
• New orders would appear to be a valid leading
  indicator but because of shorter delivery times,
  it too has become a coincident indicator.
• Similarly, improvements in inventory planning and
  control means vendor performance (delivery
  delays) is no longer relevant.
• Length of the workweek is dominated by random
  events and has little information content in the
  short run.
• That immediately eliminates (1), (3), (4), (5),
  and (8). How about the other five components?

14
Leading Indicator Components, Slide 3

• Initial claims. A valid leading indicator. But
  look at the 4-week moving average, not the actual
  weekly stats.
• Building permits. Monthly data tend to be
  distorted by weather conditions, so you need to
  look at 3 consecutive months
• Yield spread we have already shown that is an
  important and useful indicator
• Stock market and consumer expectations will now
  be examined separately.

15
Stock Market as a Leading Indicator

• Until 2001, worked quite well in anticipating
  downturns, but major declines in 1962 and 1987
  were not followed by recessions.
• In 2000, market peaked in March, but almost
  matched those highs again in September. Either
  way, that was a valid signal since the recession
  did not start until 2001.
• The upturn in October 2001 gave only a 1 month
  lead, far less than the previous average of 5 to
  6 months.
• More serious, the severe downturn in 2002 was not
  followed by another recession.
• Earlier in the previous decades, substantial dips
  in 1994 and 1998 were not followed by recessions.
  
• Scorecard over the past decade, 2 right, 3
  wrong. Use with extreme caution.

16
Consumer Expectations as a Leading Indicator

• Same general idea. Doesnt miss recessions, but
  gives too many false signals, including 1992 and
  1994.
• Double-dip in 2001 a confusing signal, although
  9/11 presumably not part of the usual pattern.
• Decline in late 2000 and sharp recovery in late
  2001 were both valid indicators.
• Worth watching, but be careful of false signals.

17
Industry Surveys ISM

• The Institute for Supply Management, formerly
  known as the National Association of Purchasing
  Managers, issues a closely followed monthly
  survey.
• It contains several components, including
  individual series for production, orders, and
  employment, but the overall index is the most
  closely followed.
• Used to emphasize trends in inflation, but
  obviously that is not a major issue lately.

18
ISM Survey, Slide 2

• Gave a valid signal for the downturn in early
  2001 and the upturn in late 2001.
• Major problem with this index is it generally
  gives very short lead time signal, or no signal
  at all. Attempts to disaggregate and use only
  the leading indicator components have not
  worked.
• Even with short lead time, worth following so you
  dont miss the recession or recovery when they do
  start.

19
Other Surveys

• Regional Federal Reserve Bank surveys of economic
  activity in those regions.
• Regional ISM surveys
• Help wanted and employment hiring surveys
• Weekly index of consumer sentiment
• For the most part, these have not passed the
  market test in the sense they are ignored by
  financial market traders because they do not
  contain enough useful information.

20
Copper Prices

• An old saying has it that Dr. Copper is a better
  predictor than 95 of PH.D. economists.
• Did a credible job of predicting the downturn and
  upturn in 2001.

21
Summary What to Follow

• 4-week moving average of initial unemployment
  claims
• Yield spread (when inverted)
• Index of consumer expectations
• ISM overall index
• Commodity prices, especially copper
• Please note list does NOT include stock market
• Good luck! Forecast at your own risk.