On-line International Price Discrimination with and without Arbitrage conditions Enrico Bachis...
-
Upload
yvonne-pooley -
Category
Documents
-
view
216 -
download
1
Transcript of On-line International Price Discrimination with and without Arbitrage conditions Enrico Bachis...
On-line International Price Discrimination with and without
Arbitrage conditions Enrico Bachis Claudio A PigaNottingham University Loughborough University
Introduction• A new form of Price Discrimination on-line• Surprisingly, this comes with Arbitrage opportunities• PD and Arbitrage opportunities persist over time• Evidence consistent with high (not low) search costs on-
line.• Discrimination seems to constitute a highly flexible pricing
strategy, suitable to very different market conditions. • Finally, discrimination is more likely within two weeks
prior to departure, suggesting it is used to manage stochastic demand.
Example 1: no PD
~118GBP
No PD: a theoretical rationale
Consistent with a situation where the airlines are confident that aggregate demand is sufficiently high to fill the flight to capacity.
Thus, the single price corresponds to the maximum fare a passenger in either country is willing to pay.
Search costs are not relevant here,
Example 2: PD and Arbitrage
Yes
It can be
Done!!
PD and Arbitrage: a theoretical rationale
The airline believes aggregate demand is not enough to fill the flight to capacity, and therefore resorts to standard third-degree price discrimination to maximize a flight’s load factor.
the high demand group is made up of passengers that are returning to their country of residence (the Britons), thereby raising the possibility of arbitrage.
Absent search costs, it is likely any price divergence would be arbitraged away. It is reasonable to assume that the presence of consumers with positive search costs makes on-line price discrimination and arbitrage a feasible strategy for the LCCs.
Example 3: PD and Arbitrage
~41GBP
Example 4: Perfect Segmentation
~83GBP
Perfect segmentation: a theoretical rationale
The airline believes aggregate demand is not enough to fill the flight to capacity, and therefore resorts to standard third-degree price discrimination to maximize a flight’s load factor.
the high demand group is made up of passengers that are leaving their country of residence (the Italians), so no possibility of arbitrage.
Discussion• An innovative form of on-line pricing, different from any
discussed in literature.• For the same flight, at the same time, the airline posts two
different fares denoted in different currency, that violate the Law of One Price.
• Price Dispersion is only due to Discrimination, as no cost difference exist between the two prices.
• Arbitrage opportunities seem to accompany some cases of Discrimination
• Surprising because arbitrage is assumed to be incompatible with discrimination (Lars Stole, 2005; Jean Tirole, 1988)
Data Collection1. Primary data on fares and secondary data on routes traffic
2. May 2002: fares collected using an “electronic spider” from only the main LCCs (i.e., Ryanair, Buzz, Easyjet, GoFly)
3. In Jan03, the “spider” was updated to retrieve fares from the Bmibaby and MyTravelLite sites. Flybe hard to crack!!!
4. Prices were gathered for both UK domestic and to European destination flights.
5. We collected the fares for departures due, respectively, 1, 4, 7, 10, 14, 21, 28, 35, 42, 49, 56, 63 and 70 days from the date of the query. No kind of study where these fares are studied for such a long period and for so many routes.
Data Collection 2
• Other flights info, such as date and time of departure and flight code, were also saved.
• Each query for a round-trip was carried out separately (but simultaneously).
• In one type of query, the outgoing flight originated in UK and fares were denoted in GBP
• In the other, the direction of the round-trip was reversed, the outgoing flight originated in continental Europe and the fares were denoted in the country of departure’s currency.
UKirtcbP
EUirtcbP
The matching of queries
UKirtcbP EU
irtcbP
Conditions for the LOP 0/ / UK
irtcbb
UKEUEU
irtcb PeP
Violations to LOP occur only for international flights, where search costs are higher.
Plus, data within Europe (not in the paper) confirms that fares in the same currency (Euro) do not differ
Company Statistic Bmibaby RyanAir EasyJet Buzz GoFly MyTravel Total International Flights p1 0.60 0.06 0.05 0.03 0.05 0.50 0.06 p5 1.23 0.39 0.24 0.23 0.29 0.91 0.32 p10 1.67 0.88 0.46 0.45 0.62 1.37 0.63 p25 2.56 2.32 1.07 1.03 1.91 2.53 1.50 p50 4.25 5.32 2.35 2.05 3.62 4.20 3.41 p75 7.15 9.93 4.15 3.45 9.56 6.32 6.53 p90 10.67 17.20 5.92 8.17 16.95 10.13 12.13 p95 14.58 23.51 8.53 14.01 23.23 14.35 17.50 p99 22.79 36.81 17.58 29.50 42.21 32.96 34.08 mean 5.56 7.68 3.17 3.65 7.05 5.53 5.38 min 0.00 0.00 0.00 0.00 0.00 0.01 0.00 max 79.99 79.84 79.97 79.91 79.98 79.36 79.99 sd 4.85 8.08 3.81 5.62 8.65 5.94 6.53 N 168750 803782 849313 42333 30957 23289 1918424
Domestic Flights p1 0 0 0 - 0 0 0.00 p5 0 0 0 - 0 0 0.00 p10 0 0 0 - 0 0 0.00 p25 0 0 0 - 0 0 0.00 p50 0 0 0 - 0 0 0.00 p75 0 0 0 - 0 0 0.00 p90 0 0 0 - 0 0 0.00 p95 0 2.52 0 - 0 0 0.00 p99 0 10 5.00 - 10 0 5.00 mean 0.05 0.40 0.12 - 0.26 0.00 0.18 min 0 0 0 - 0 0 0.00 max 55.50 63.00 75.00 - 39.15 3.99 75.00 sd 0.88 2.05 1.39 - 2.20 0.13 1.54 N 54601 71408 137083 - 7534 1772 272398 Total N 223351 875190 986396 42333 38491 25061 2190822
Violating the LOP - 2b
UKEUUK
irtcbEU
irtcb ePP //
05
1015
2025
Ke
rnel
de
nsity
1.2 1.4 1.6 1.8
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of Ryan Air
05
1015
20K
ern
el d
ens
ity
1.3 1.4 1.5 1.6 1.7
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of EasyJet
010
2030
Ke
rnel
de
nsity
1.4 1.45 1.5 1.55 1.6 1.65
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of BmiBaby
010
2030
40K
ern
el d
ens
ity
1.4 1.5 1.6 1.7 1.8
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of MyTravelLite
020
4060
8010
0K
ern
el d
ens
ity
1.4 1.5 1.6 1.7
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of GoFly
05
1015
20K
ern
el d
ens
ity
1.3 1.4 1.5 1.6 1.7
Ratio EUPrice/GBPrice Exchange Rate
Euro-area routesKernel Density of Buzz
A Taxonomy of cases UK
irtcbb
UKEUEU
irtcb PeP //
Discrimination Type values
Condition Logic
Oper.
Condition
0 – Non Discriminatory.
|Δ|<5
1- Discrimin. no arbit.
(UK AND Δ?-5) OR (Cont.EU AND Δ?5)
2 – Discrimin. No prof arbit.
(UK AND Δ?5 AND Δ?AC) OR (Cont.EU AND Δ?-5 AND Δ?-AC)
3– Discrimin. with prof arbit.
(UK AND Δ?5 AND Δ?AC) OR (Cont.EU AND ?-5 AND Δ?-AC)
)/,min(05.00.6 bEUirtcb
UKirtcb ePPAC
Is arbitrage rare? Variable “Discrimination Type” – Frequency (row %)
Departure From
0- Non discriminatory
1 Discriminatory - no arbitrage
2 Discriminatory -no prof. arbitrage
3 Discriminatory - with arbitrage
Cont. Europe 49534 (58.9%) 34357 (40.8%) 18 (0.0%) 211 (0.3%) UK 51281 (60.6%) 112 (0.1%) 24912 (29.4%) 8325 (9.8%)
Bm
i Bab
y
Total 100815 (59.7%) 34469 (20.4%) 24930 (14.8%) 8536 (5.1%) Cont. Europe 193133 (48.1%) 62909 (15.7%) 36163 (9.0%)
109552 (27.3%)
UK 191923 (47.7%) 143864 (35.8%) 20756 (5.2%) 45482 (11.3%)
Rya
n
Air
Total 385056 (47.9%) 206773 (25.7%) 56919 (7.1%) 155034 (19.3%)
Cont. Europe 329673 (78.7%) 77628 (18.5%) 7665 (1.8%) 4143 (1.0%) UK 381208 (88.6%) 16873 (3.9%) 27315 (6.3%) 4808 (1.1%) E
asy
Jet
Total 710881 (83.7%) 94501 (11.1%) 34980 (4.1%) 8951 (1.1%) Cont. Europe 17673 (84.2%) 2463 (11.7%) 169 (0.8%) 679 (3.2%) UK 18437 (86.4%) 1221 (5.7%) 767 (3.6%) 924 (4.3%) B
uzz
Total 36110 (85.3%) 3684 (8.7%) 936 (2.2%) 1603 (3.8%) Cont. Europe 9317 (60.6%) 2170 (14.1%) 1636 (10.6%) 2240 (14.6%) UK 9595 (61.5%) 3624 (23.2%) 504 (3.2%) 1871 (12.0%) G
o
Fly
Total 18912 (61.1%) 5794 (18.7%) 2140 (6.9%) 4111 (13.3%) Cont. Europe 6657 (61.0%) 4021 (36.9%) 62 (0.6%) 167 (1.5%) UK 7762 (62.7%) 188 (1.5%) 3244 (26.2%) 1188 (9.6%) M
TL
Total 14419 (61.9%) 4209 (18.1%) 3306 (14.2%) 1355 (5.8%) N 1266193 349430 123211 179590 %N 66.0% 18.2% 6.4% 9.4%
For Britons
For EU travellers
EU travellers adversely
discriminated
Non profitable Arbitrage for EU travellers
Is Arbitrage worthwhile? Variable “Discrimination Type”
Departure From
0- Non discriminatory
1 Discriminatory - no arbitrage
2 Discriminatory -no prof. arbitrage
3 Discriminatory - with arbitrage
Mean 2.9 9.7 7.1 15.9
Arbitrage Cost 7.7 10.1 9.3 11.3
Bm
i Bab
y
Mean UK
irtcbP 33.5 82.1 66.6 107.5
Mean 2.3 12.8 6.4 14.8
Arbitrage Cost 7.2 7.6 7.8 7.5
Rya
n A
ir
Mean UK
irtcbP 24.6 41.0 40.7 40.5
Mean 2.1 8.4 7.0 19.2
Arbitrage Cost 8.0 10.1 10.6 8.8
Eas
y Je
t
Mean UK
irtcbP 41.0 83.0 93.7 65.3
Mean 1.9 13.6 6.4 19.2
Arbitrage Cost 8.1 8.7 10.0 7.2
Buz
z
Mean UK
irtcbP 42.9 58.8 80.6 32.5
Mean 2.3 14.7 7.3 18.2
Arbitrage Cost 9.0 9.2 10.6 9.0 Go
Fl
y
Mean UK
irtcbP 61.3 73.6 98.1 70.6
Mean 2.9 9.7 6.9 17.4
Arbitrage Cost 7.1 9.3 9.0 10.3
MT
L
Mean UK
irtcbP 22.5 66.5 60.5 89.3
E-commerce• A typical assumption in e-commerce literature:• Low search costs lead to high price transparency.• Solution to this: Brand allegiance and/or obfuscation
strategies.• The existing literature assumes away the possibility of
different prices for the same e-retailer, on the same site at the same time.
• Price Dispersion is across e-retailers, both in theoretical and empirical work.
• Theoretical reference for Dispersion within the same firm: Salop (1977, RevEcStud)
Making arbitrage consistent with on-line PD
• Two non-mutually exclusive explanations for our findings:
1. In a particular sense, Search costs may not be as low as one might think:
• bounded rationality due to psychological inertia prevents the search in the first place.
2. The airlines may not be too worried if some arbitrage is exercised, when demand is low.
The econometric model• We use a Bivariate Probit model with Sample
Selection to study two discrete variables:
• “Discriminatory”, and
• “Arbitrage”. This is relevant only for the sub-sample where “Discriminatory”=1
D
ND
A
NA
Note: NA includes non-profitable arbitrage cases
Descriptives
Marginal Effects
Estimates
Econometric Results• Persistence over time characterizes discriminatory
cases • arbitrage opportunities may remain posted for long
periods• evidence generally suggests a relation between
discriminatory observations and route concentration • “Market Size” and the presence of charter operators
are both positively correlated with discriminatory cases
Conclusions
• We conclude that PD seems to be of a competitive type.
• arbitrage opportunities tend to be more likely when the airlines enjoy a dominant position.