Biography for William Swan

1
Biography for William Swan
Chief Economist, Seabury-Airline Planning Group.
AGIFORS Senior Fellow. ATRG Senior Fellow.
Retired Chief Economist for Boeing Commercial
Aircraft 1996-2005 Previous to Boeing, worked at
American Airlines in Operations Research and
Strategic Planning and United Airlines in
Research and Development. Areas of work included
Yield Management, Fleet Planning, Aircraft
Routing, and Crew Scheduling. Also worked for
Hull Trading, a major market maker in stock index
options, and on the staff at MITs Flight
Transportation Lab. Education Masters,
Engineers Degree, and Ph. D. at MIT. Bachelor
of Science in Aeronautical Engineering at
Princeton. Likes dogs and dark beer.
(bill.swan_at_cyberswans.com)

• Scott Adams

2
Boeing 11-Year World Airline Traffic
OutlookShort Form of what we do for a living in
my shop
Growth is an Average Over Cycles
GDP growth for world 2.8 RPK
growth for the world 4.4 We
predict these averages
Tables, regional growth, charts, lots of stuff on
resultant aircraft fleets http//www.boeing.com/c
ommercial/cmo
3
Underlying Theme

• A TREND is a projection of past growth
• A FORECAST includes reasons why
• The FUTURE includes acts of will

4
My Forecast Could Have been a Trend
5
Three Forecasting Mistakes

• We now have a new forecasting method
• We still have similar results

William M. SwanChief EconomistBoeing Marketing
6
Three Lessons (our mistakes)

• Do not let the models form determine the answer
• -- algebra can force interpretation
• Statistics do not preclude the use of reason
• --logical demonstrations are also valid
• Do not give up
• --statistics can see through scatter
• Plus our Bonus lesson
• An example of business sleaze(Occams
  Toothbrush)

7
Do Not Let the Models Form Determine the Answer

• Our traditional formula
• RPKs GDP? Yield-?
• ? is the GDP elasticity
• ? is the price elasticity

• Problems with algebra
• GDP did not distinguish between sources of growth
• -- per-capita wealth increase GDP or population
  growth GDP?
• Yield is a poor surrogate for ticket prices
• All growth MUST be attributed to GDP or Yield

8
Error 1 Travel Does Not Grow as GDP1.5

• Cross-sectional data Travel Share not rising
  with incomes
• Travel Share of GDP measured as ASK/GDP ratio
• Data shows small negative correlation with
  per-capita income
• No acceleration of travel share after joining
  middle class
• Time-series data confirmed pattern
• Growth of Travel Share was independent of growth
  of GDP
• Based on Country-by-Country data
• Conclusion
• Travel grows linearly with GDP growth
• Remaining 1/3 of travel growth is something
  else
• Useful question What Else?

9
Air Travel Share of GDP Is Independent of Income
10
Independent Data Source Same Pattern
11
Same Data, Different Display No S in Travel
Per Person
12
Same Data, Different Display Say Good-Bye to the
S-Curve
13
GDP Elasticity Absorbed All Growth
Lousy data on fares price elasticity
understated. Therefore, GDP picked up price
effect RPK(t) GDP(t)? Yield(t)-? became
RPK(t) GDP(t)?
but should have been
ASK/GDP GDP0 F(t) Travel Share (ASK/GDP)
grows with time (t) Time trend F(t) confounded
earlier analyses because there wasnt one in
the algebra
14
Another Reason GDP Was Over-Emphasized Beware
Heteroscedasticity

• Traffic in short term responds to consumer
  confidence
• Consumer confidence moves when GDP moves are
  large
• Therefore large GDP moves show large traffic
  responses
• Least-squared calibration weights large moves
  more
• Result
• Calibrations over-stated GDP elasticity of travel
• Long-term growth must expect average consumer
  confidence

15
The What Else? question Time trend is other
effects

• Price effects
• More nonstops stimulate travel
• Global trade stimulates travel

Secular Trend in Travel Share of GDP
16
Yield Overstates Fare Declines Yield is an
Imperfect Statistic

• Yield is an average
• Average yield declines with more long trips
• Average yield declines with more discount
  (pleasure) trips
• Under half of 2.2 yield decline is decline in
  fares
• Business fares have gone up
• Pleasure fares have gone down, and quality to
  match

17
Humility Measurement Biases Half the time an
Economist talks, he talks about measures, not
answers

• Taxes and airport fees seem to keep going up
• Added 0.4 /year to ticket price in US last
  decade
• Greater increases for international flying
• Reported yields are net to the airline
• Actual fare changes had tax and fee increases
• The other messy nit inflation
• In the US, CPI inflation is 0.3 higher than GDP
  inflation
• CPI overstates inflation
• 0.3 growth is not negligible
• In the UK, similar answer but smaller difference
• Elsewhere, not so bad

Please do not compare GDP growth using GDP
deflator with airline revenue growth without
taxes deflated by CPI
18
Hard Lesson Work the Data

• It is easy to grab data a run with it
• It takes work to muck about in the data
• Learn what is being measured
• Play with the data, examine outliers
• You will gain more from good data than good
  modeling
• Professors only publish modeling

19
Error 2 Value of Service Can be MeasuredIt
has proven almost impossible to calibrate service
elasticities But that does not preclude the use
of reasoning Here is the reasoning

• ASKs doubled with almost no growth in aircraft
  size
• Lots of new flights, new times, new places
• It could have gone the other way
• Just bigger airplanes would have saved 1.5/year
• Cost Savings foregone represent value added by
  new services
• Market produced high-service result
• Implication is that value exceeds cost savings
• Surprise! Value growth exceeds fare reductions
  in size

20
Value of Better Service Approximated by Cost
Growth 1985-1995 Schedules

• Growth absorbed almost entirely with frequencies.
• Foregone cost savings approximately 15 in 10
  years.
• Value created approximately 15 in 10 years
• This should stimulate as much travel as a fare
  decrease

21
Humility Service got WorseHalf the time an
Economist talks, he talks about measures, not
answers

• We estimated costvalue of more direct services
  and frequencies
• We did not estimate the loss of value associated
  with
• Lower reliability
• More delays
• Smaller personal space on airplane
• Higher load factors
• Worse food
• Busier flight attendants
• Worse airport access
• Longer airport processing times
• More trouble finding the best fare
•  So we may have overstated the net quality
  effect.

22
Error 3 Scatter is Not NoiseDo not give up
the statistics
23
Cheating on Statisticswhat explains scatter?

• Quality of cuisine hypothesis -- untasted
• A Statistically Significant result is at 95
• Means 5 chance of being random coincidence
• Ran 40 regressions, found 2 Significant results
• women in the workforce
• Politically incorrect, do not pursue
• International trade as of GDP
• Tells a good story, makes sense, follow this up
• Moral of the story statistics can find new ideas
• If in publish or perish world
• Write two papers, get two brownie points

24
Travel Grows With TradeInternational Trade
drives some Air Travel Growth

• Travel Share (ASK/GDP Ratio) grows with increased
  Trade
• Trade measured as ImportsExports as of GDP
• Trade growing nearly twice as fast as GDPs
•  Cross-sectional data significant
• Trade explains some of the scatter in Travel
  Share
• Time-series data also significant
• Change in Trade creates change in Travel
  Service\
• Same ratios, either way

25
Bonus Lesson An example of business sleaze
Proof by Assumptions Test (Occams
Toothbrush) Introducing a technique often used
in business
26
Occams Toothbrush
Is there a reasonable set of assumptions that
fit all known data AND Allow my answer to be
right?
27
Final Surprise Business Demand is growing
almost as fast as
Pleasure Demand (Demand is the demand curve,
not the traffic count)

• Proof by Occams Toothbrush
• What is the most reasonable set of assumptions
  that allow data?
• Business travel share (survey data) declines only
  3 in 10 years
• We expect trade growth to drive business travel
• We expect service quality ltgt business travel
• Overall traffic has been growing 5 annually
• Fares are declining only for pleasure demands
• What is minimum business demand growth beyond the
  3 from GDP?
• -- it turns out pretty high

28
Business Share of Traffic Declines SlowlyWhat
set of assumptions fits all the available data?
29
Conclusion Data is a NuisanceContinually
upsetting well-established platitudes

• Nobody can tell if a forecast is right
• Everybody can tell if a forecast has changed
• We have not changed total trends of growth
• We have changed
• Where in the world the growth may be
• What we look for if trends are to change
• How we explain it

It is better to light one poor candle than to
curse the darkness.
30

31
William Swan
Data Troll Story Teller Economist
32
(No Transcript)