The use of time series analysis for the analysis of airlines

The use of time series analysis for the analysis of airlines

D.E.PitfieldTransport Studies Group

Department of Civil and Building EngineeringLoughborough University

LoughboroughLeicestershire LE11 3TU

UK

Paper presented at Fifth Israeli/British & Irish Regional Science Workshop, Ramat-Gan, Tel-Aviv, Israel, 29 April - 1 May 2007.

• Time Series Applications

– Oligopolistic Pricing of Low Cost Airlines• Cost Recovery?

– Impact of Ryanair on Market Share and Passenger Numbers

– Impact of Airline Alliances?• formation• Open skies agreements

Figure 1: A Location Map of Nottingham East Midlands Airport, UK.

Source: http://www.multimap.com/

DAY

29.0025.0021.0017.0013.009.005.001.00

£s 100

80

60

40

20

0

bmibaby fare

easyJet fare

Figure 3: Fares from EMA to Alicante

DAY

29.0025.0021.0017.0013.009.005.001.00

£s 100

80

60

40

20

0

bmibaby fare

easyJet fare

Figure 4: Fares from EMA to Malaga

Figure 15: Fares from LGW to Prague

DAY

49454137332925211713951

£s120

110

100

90

80

70

60

50

40

bmibaby fare

easyJet fare

Figure 7: CCF plot: Malaga

Lag Number

7531-1-3-5-7

CC

F -

bm

ibaby

and e

asyJ

et

1.0

.5

0.0

-.5

-1.0

Confidence L imits

Coefficient

ACF: bmibaby 0.899 easyJet 0.650

• ACF bmibaby 0.899 easyJet 0.650

CCF: 0.452 at lag 1day easyJet leading bmibaby

Figure 10: CCF plot: Alicante

Lag Number

7531-1-3-5-7

CC

F -

bm

ibaby

and e

asyJ

et

1.0

.5

0.0

-.5

-1.0

Confidence L imits

Coefficient

CCF: 0.808 at Lag 0

ACF:

bmibaby 0.375 easyJet 0.535

Figure 18: CCF plot. LGW-PRA

Lag Number

7531-1-3-5-7

CC

F -

bm

ibab

y a

nd e

asyJet

1.0

.5

0.0

-.5

-1.0

Confidence Limits

Coefficient

Figure 1: Ryanair’s Route Network

Figure 2: London Area Airports

Selected Airports

• Genoa

• Hamburg

• Pisa

• Stockholm

• Venice

London-Venice 1991-2003

JAN

1991A

UG

1991M

AR

1992O

CT

1992M

AY

1993D

EC

1993JU

L 1994F

EB

1995S

EP

1995A

PR

1996N

OV

1996JU

N 1997

JAN

1998A

UG

1998M

AR

1999O

CT

1999M

AY

2000D

EC

2000JU

L 2001F

EB

2002S

EP

2002A

PR

2003N

OV

2003

Date

0.00

10000.00

20000.00

30000.00

40000.00lgw

lhr

lcy

stntsf

stnvce

London-Venice 1991-2003

1999

18.5%

37.4%21.9%

22.2%LGW

LHR

LCY

STN-TSF

STN-VCE

2000

24.5%

25.7%28.6%

21.2%LGW

LHR

LCY

STN-TSF

STN-VCE

2001

45.5%

37.3%

17.2% LGW

LHR

LCY

STN-TSF

STN-VCE

2002

33.3%

45.6%

21.1%LGW

LHR

LCY

STN-TSF

STN-VCE

2003

30.8%

6.3%43.2%

19.6%LGW

LHR

LCY

STN-TSF

STN-VCE

Venice Intervention Model - with regular differencing

 Parameters t tests Goodness of Fit 

MA1 0.565 8.019 SE = 0.084 

SAR1 -0.458 -5.981 Log Likelihood = 151.540 

Intervention Ryanair

0.258 4.548 AIC = -295.081 

Intervention GO

0.236 4.165 SBC = -283.229 

RMS= 3156.129 U = 0.037 Um = 0.003, Us =0.001, Uc = 0.995

Minimum Start-Up Impact of Ryanair by destination

• Genoa – 44%

• Hamburg – 12%

• Pisa – 30%

• Stockholm – 10%

• Venice – 26%

Alliances• Oum et al (2000) Globalization and Strategic

Alliances: The Case of the Airline Industry

– Parallel Alliances

• Competition decreases

• Coordination of schedules

• Restricted output

• Increased fares

• FFPs

– Complementary Alliances

• Fares fall• Network Choices Improve• Traffic Falls?• Alliance Share increases?

Expectations and Perceptions

• Iatrou, K & Alamdari, F. (2005), The Empirical Analysis of the Impact of Alliances on Airline Operations, Journal of Air Transport Management

• Impact on traffic and shares is positive– hubs at O and D?– 1-2 years – Open skies has biggest impact

Data• North Atlantic – scale and role of alliances

• BTS T-100 International Market Data– monthly, January 1990- December 2003

• Hubs– Choice?

• European – LHR, CDG, FRA, AMS– not LHR or AMS

• USA – JFK, ORD, LAX

• Parallel

– CDG – JFK (Skyteam – AF and DL)– FRA – ORD ( Star Alliance – LH and UA)

• Complementary

– FRA – JFK ( Star Alliance – LH)– FRA – LAX (Star Alliance – LH/NZ)– CDG/ORY – BOS (Skyteam – AF)

ARIMA and Intervention Analysis

• Model traffic before Intervention(s)– Using parsimonious models

• Specify Intervention term and model whole data series– Abrupt impact– Gradual impact, over one or two years

• Exponential or stepped

– Lagged Abrupt impact

Figure 4.1: Traffic CDG-JFK 1990-2003

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

25000.00

20000.00

15000.00

10000.00

5000.00

0.00

To

tal

Figure 4.11: Alliance Share, CDG-JFK 1990-2003

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

100.00

80.00

60.00

40.00

20.00

0.00

AF

shar

e

Paris (CDG) – New York (JFK)

A B C

Average monthly Average monthly Average monthly

traffic in the quarter traffic in the quarter traffic in the quarter

including start 1 year after A 2 years after A

of intervention

Traffic

Code sharing

42,573 54,529 58,128

Immunity

33,290 32,817 36,339

Alliance Share %

Code sharing

73.2 72.1 71.1

Immunity

77.9 77.4 75.8

• Seems? Traffic stimulated after code sharing and immunity. Shares?

• Intervention Analysis? – no significant intervention. Indigenous influences on traffic more important as well as other exogenous influences i.e. ceteris paribus

including 9/11 – 42% drop in total

Figure 4.2: Traffic CDG/ORY-BOS 1990-2003

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

25000.00

20000.00

15000.00

10000.00

5000.00

0.00

To

tal

Figure 4.21: Alliance Share, CDG/ORY-BOS 1990-2003

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

100.00

80.00

60.00

40.00

20.00

0.00

AF

shar

e

Paris (CDG/ORY) – Boston (BOS)

A B C

Average monthly Average monthly Average monthly

traffic in the quarter traffic in the quarter traffic in the quarter

including start 1 year after A 2 years after A

of intervention

Traffic

Code sharing

12,858 13,481 14,767

Immunity

10,434 8,924 10,004

Alliance Share %

Code sharing

47.2 61.7 69.8

Immunity

65.2 100.0 100.0

• Seems? Traffic increased from code sharing but not immediately from immunity. Shares? – AA!

• Intervention? Only nearly significant results are of a negative impact for traffic!

But this reflects 9/11 impact– Cannot model shares as partners have 0

traffic for some months

Figure 4.3: Traffic FRA-JFK 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

60000.00

50000.00

40000.00

30000.00

20000.00

10000.00

To

tal

Figure 4.31: Alliance Share, FRA-JFK 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

80.00

60.00

40.00

20.00

0.00

LH

UA

shar

e

Frankfurt(FRA) – New York(JFK)

A B C

Average monthly Average monthly Average monthly

traffic in the quarter traffic in the quarter traffic in the quarter

including start 1 year after A 2 years after A

of intervention

Traffic

Code sharing

42,064 42,856 43,090

Immunity

40,623 29,872 32,630

Alliance Share %

Code sharing

30.6 32.7 32.5

Immunity

33.0 46.5 51.7

• Seems? Little impact on traffic but impact on shares

• Intervention – not significant apart from a possible negative impact-contradicts expectations and theory of

complementary alliances

Figure 4.4: Traffic FRA-ORD 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

40000.00

30000.00

20000.00

10000.00

To

tal

Figure 4.41: Alliance Share, FRA-ORD 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

90.00

80.00

70.00

60.00

50.00

LH

UA

shar

e

Frankfurt (FRA) – Chicago (ORD)

A B C

Average monthly Average monthly Average monthly

traffic in the quarter traffic in the quarter traffic in the quarter

including start 1 year after A 2 years after A

of intervention

Traffic

Code sharing

17,889 21,030 22,392

Immunity

22,392 23,632 32,472

Alliance Share %

Code sharing

73.1 74.5 76.8

Immunity

76.8 79.4 83.5

• Seems? Alliance partners hub at origin and destination so may expect a positive impact

• Traffic seems to increase especially from open skies. Shares up at both interventions

• Intervention. Results are positive and nearly significant contrary to theory of parallel alliances. Best results but not conclusive.

Figure 4.5: Traffic FRA-LAX 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

25000.00

20000.00

15000.00

10000.00

5000.00

To

tal

Figure 4.51: Alliance Share, FRA-LAX 1990-2003

JAN

200

5

JAN

200

4

JAN

200

3

JAN

200

2

JAN

200

1

JAN

200

0

JAN

199

9

JAN

199

8

JAN

199

7

JAN

199

6

JAN

199

5

JAN

199

4

JAN

199

3

JAN

199

2

JAN

199

1

JAN

199

0

Date

100.00

90.00

80.00

70.00

60.00

50.00

40.00

LH

UA

shar

e

Frankfurt (FRA) – Los Angeles (LAX)

A B C

Average monthly Average monthly Average monthly

traffic in the quarter traffic in the quarter traffic in the quarter

including start 1 year after A 2 years after A

of intervention

Traffic

Code sharing

14,511 18,264 18,622

Immunity

18,622 19,319 17,134

Alliance Share %

Code sharing

51.1 54.4 51.4

Immunity

51.4 74.4 83.7

• Seems? Traffic stimulated from code sharing and shares up from open skies

• Intervention – no significant results. Major impact is probably the withdrawal of Continental some 11 months later and this causes alliance share to grow

Conclusion• Weak evidence suggests that impact of

complementary alliances is to reduce traffic and shares. Contrary to all theory.

• Some evidence that positive impact from parallel alliances when participants hub, but this is contrary to theory cf. expectations.

Generally, other things matter.

• Open Skies agreements appear to cause a decrease in traffic and competition; true for all alliance types – transatlantic traffic may not grow as these agreements spread.

• Alliance strength may be barrier to entry