[1] Council regulation No.EEC/3975/87 (O.J.1987 L 374/1).
[2] Council regulation No.EEC/3976/87 (O.J.1987 L 374/9).
[3] Council Directive No.87/601/EEC (O.J:1987 L374/12).
[4] Council regulations No.EEC/2342/90 and No.EEC/2343/90 (O.J.1990 L 217/1 and 8).
[5] Council regulations No.EEC/2407/92 and No.EEC/2409/92 (O.J.1992 L 240/1 and 15).
[6] In such cases, the member state may ask the Commission to examine the fare and decide whether or not the Commission can approve it (Council Regulation No.EEC/2409/92).
[7] In Council Regulation No.EEC/2407/92 (O.J.No.1992 L 240/1).
[8] Council Regulation No.EEC/2408/92.
[9] See Council Regulation No.EEC/2408/92. The reason why cabotage was not liberalized completely until 1 April 1997 was strong resistance from the larger EC countries, which at the time when the third package came into operation were not ready to open their domestic markets and to allow competition on busy routes.
[10] The following characterizations of each school are to a great extent shortened reviews from Gilbert's introduction in Geroski, Gilbert and Jacquemin (1990)
[11] As with economies for sales promotion, advertising, etc., Bain sees no "unique and irreversible general relationship between unit sales promotion and long-run rate of output or sales". Unlike production and distribution unit costs, which are driven by the amount produced, advertising unit costs depend on the amount that can be sold over a given time or on the selling price. For Bain, promotion and advertising cost advantages remain inherently different from production and distribution cost advantages due to scale. He treats this issue as part of an analysis of product differentiation.
[12] When entering at a size smaller than minimum efficient scale, a newcomer incurs the disadvantage of higher unit costs but gains the advantage of adding a smaller output increment and thus precipitating a milder price decline. This advantage is increased if demand is elastic (the market readily absorbs more supply without a large fall in price).
[13] In contrast, with absolute cost advantages, these marginal returns would remain constant.
[14] The authors go on to describe drivers such as:
Aircraft type: The choice among different types of aircraft essentially involves different production technologies with differing input intensities and relative efficiencies. For example, wide-bodied aircraft can serve densely trafficked long-haul routes efficiently, while smaller aircraft can serve less densely trafficked short-haul routes relatively efficiently. Aircraft choices depend on network characteristics, particularly route length and traffic density, and the availability of particular models of aircraft. The output capacity of each aircraft, in turn, determines the required hours of pilot, co-pilot, flight engineer, navigator and flight attendant labour, required levels of maintenance and required quantities of fuel per capacity seat miles.
Aircraft size and average stage length: In air transportation services, output capacity increases with both the number of seats made available and the distance travelled. If the volume of traffic is heavy enough for a carrier to use larger aircraft on a given flight and route, more capacity seat miles can be delivered for a given level of flight crew labour and fuel costs.
Density: Here, the authors define density as increasing the number of flights over its network by which the carrier can offer a more diversified set of services. They see the proliferation of flights as coming about for two reasons: (1) efforts to provide a more attractive schedule of flights and (2) efforts to utilize productive capacity more fully. The conventional wisdom is that a carrier can utilize inputs more efficiently by operating more flights or carrying more traffic over a given network. There is some empirical evidence that economies of density obtain. However, the underlying dynamics in terms of increased production complexity are not addressed. By scheduling more flights over a given network segment, a carrier incurs additional set-up costs for each additional flight, in terms of handling aircraft on the ground and enplaning and deplaning passengers and cargo. At the same time, it may be able to better utilize its fixed ground property and equipment and general overhead inputs.
Hub concentration: Carriers can thus achieve substantial economies, for example, maintaining and repairing their fleets, by using ground property, equipment and labour, and by filling larger aircraft on hub to hub routes. However, to achieve these economies, it is likely that carriers will also have to use more administrative and supervisory labour for communications and other support services. Moreover, at major hubs there may be intense competition for the use of air traffic control and of shared ground facilities. This often results in overscheduling, congestion and delays. The magnitudes of the economies a carrier can obtain by concentrating flights through hubs are likely to depend on whether the carrier has some monopoly power, as reflected in the dominant market shares at its hub airport.
Scale: In a cost accounting framework of constant marginal cost, increasing returns to scale are indicated when fixed costs are present because average costs decrease with increasing levels of outputs. Therefore, ground property and equipment, general overhead, maintenance labour, maintenance materials and overhead inputs, all of which have fixed cost components, are likely to have increasing returns to scale associated with them, but constant returns to scale are likely to obtain for other cost categories.
[15] PORTS = number of airports served, ADPP = average number of departures per port, ASL = average stage length, AAS = average aircraft size, ALF = average load factor.
[16] Returns to density were defined as the proportional increase in output made possible by a proportional increase in all inputs, with the number of airports served, average stage length, average load factor and input prices held fixed.
[17] Returns to scale were defined as the proportional increase in output and airports served made possible by a proportional increase in all inputs, with ASL, ALF and input prices held fixed.
[18] Such economies would mean that more densely travelled spokes might have lower marginal costs. This would lead to economies of scope across itineraries that share a common spoke. These economies of scope in turn imply network economics.
[19] Their marginal cost function would be: MC = 110.9 + 0.062 * DIST - 0.00117 * Q, where Q is the spoke traffic. In a high density network, MC would equal $107, in a low density network, MC would be $134 and in a network with moderate densities, MC would be $121.
[20] see Kirby 1986, p. 346: Table 3 for an estimated cost elasticity with respect to changed composition of output.
[21] MC = 110.9 + 0.062 * DIST - 0.00117 * Q, with Q being the spoke traffic.
| DIST / DENS | 0.25 | 0.50 | 0.75 | 1.00 | 1.50 | 2.00 |
| 250 miles | 0.794 | 0.808 | 0.733 | 0.568 | -0.27 | -0.980 |
| 500 miles | 0.125 | 0.204 | 0.194 | 0.096 | -0.369 | -1.190 |
| 1,000 miles | -0.737 | -0.527 | -0.406 | -0.373 | -0.576 | -1.135 |
| 1,500 miles | -0.964 | -0.623 | -0.371 | -0.207 | -0.148 | -0.446 |
| 2,000 miles | -0.556 | -0.084 | 0.299 | 0.593 | 0.914 | 0.879 |
[23] Empirical data suggest that incumbents' advantages are more likely to originate from higher prices than lower selling costs than entrants: Advertising and distribution costs as a percentage of total operating expenditures are shown for different carriers in the UK (Civil Aviation Authority 1998, p.142).
| Distribution costs 1996 | BA | BM | Air UK | Ryanair |
| (% of operating costs) | 29% | 20% | 20% | 18% |
[24] Civil Aviation Authority (1998) also points out that the incumbents increased flight frequency on the examined routes. Therefore, it cannot be assumed that market share was necessarily lost by charging higher prices than entrants.
[25] The average net advantage due to differentiation would not be expected to remain for ever. It is likely that an entrant would anticipate large initial disadvantages, which decrease with the passage of a certain number of years.
[26] See exhibit 19
[27] "Sufficient" here meaning to provide overall positive net earnings.
[28] Refer to chapter IV.3.1.3.
[29] Tickets are considered flexible when they can be exchanged at least once against other flights without additional cost. Ticket restriction means that a flight change would incur fees either for a new ticket or for cancellation.
[30] We checked that these special fares do indeed reflect significantly lower prices compared with other ticket classes, including restricted tickets.
[31] The three elasticities are: response of costs to physical output, response of prices to differentiation- enhancing measures - i.e. advertising - and the response of prices to output.
[32] This may be the reason for the proliferation of new classes and increased price discrimination between them: Demand elasticities may become more elastic the better the market is segmented.
[33] The products are substitutes if c>0. Each firm has a cost function including fixed costs.
[34] Although Dixit ignores in this comparison the cost of product differentiation.
[35] This concept of differing degrees of differentiation is also compatible with Hotelling's model of linear product differentiation (see Tirole).
[36] We reiterate that the impact of FFPs was strongly dependent on network size and a central hub.
[37] This sheds new light on the hub-spoke advantage of airlines: The argument of economies of density, which has been put forward so often in the literature may only be secondary - if significant at all - in the European context.
[38] We assume the same elasticity of demand for both incumbent and entrant, as do Farrell and Shapiro.
[39] When we examine aircraft size, we assume the plane already to be in the air. That is, higher operating costs for maintenance, crew, fuel, depreciation, etc. do matter on an average cost basis, but not for marginal costs.
[40] The legendary price wars among the major US airlines especially during the mid 1980's display this reasoning based upon marginal cost-pricing.
[41] NB: The fact that such direct routes may originate or lead to hub airports does not imply by definition that economies of a hub-and-spoke system are exploited. If a businessman commutes from Berlin to Frankfurt, the potential for economies due to hub-and-spoke operations remain unexploited (economies of centralization, which could easily be exploited by other, long-distance routes due to the near-congestion of such hub airports).
[42] We chose to exclude all routes involving the incumbent's hub airport in order to isolate direct flights, which do not operate on a hub-and-spoke basis.
[43] See Chapter 1 for the potential efficiency gains due to hub-and-spoke operations.
[44] The table shows sample data and is in no way meant to represent exhaustively all hub or non-hub traffic flows of an incumbent. The distinction between hub and non-hub traffic only involves the incumbent's own hub, because traffic on competitors' hubs will not impact on the incumbent's marginal costs (i.e. a LH flight from Berlin to Madrid might continue with Iberia).
[45] See, for example, Civil Aviation Authority 1998, p. 142, table E12 for a comparison of cost positions per unit between BA and entrants.
[46] SAS became part of the consortium in the mid-'90s, but had dropped out of it by 1997.
[47] Amadeus web-page January 1999.
[48] In 1996 : United Airlines (38%), BA (14.7%), Swissair (13.2%), KLM (12.1%), USAirways (11%), Alitalia (8.7%), Olympic (1%), Air Canada (1%), Aer Lingus, Austrian and Air Portugal (0.1% each).
[49] Information from Amadeus and Galileo web pages January 1999.
[50] GAO report T-RCED-88-62, p. 9
[51] Doganis 1991, p. 277.
[52] For example, in the US, Continental Airlines declined to pay booking fees to PARS, one of the smaller CRS, for six weeks. Continental's decline in revenues was so great that it was forced to resume paying the fees.
[53] Of course, each of the airlines would eventually have had to operate the flight, but arrangements such as blockspacing also committed the non-operating airline to provide a certain capacity for a given city pair.
[54] These regional partnerships, though pervasive, mostly served low-density routes.
[55] From Exhibit 37 we can derive approximately the number of regional partners for the European incumbents in 1997: Aer Lingus (1), Air France (7), Air Portugal (4), Alitalia (8), Austrian (1), BA (9), Finnair (1), Iberia (3), KLM (7), Lufthansa (10), Luxair (1), Olympic (0), Sabena (6), SAS (7).
[56] Non-exhaustive, see Exhibit 37.
[57] CRS are a helpful tool in listing code-shared flights before other ones and thus biaising the client's or travel agent's choice towards the interlined carriers.
[58] Another form of excess capacity would be the interlined carrier serving given routes more frequently due to the "assured" contingent of interlined passengers.
[59] Gilbert quotes the example of taxi medallions: As strategic factors of production, they are made available to only a contingent number of taxi cabs. They may be traded, thus opportunity costs exist. However, they are not a source for supra-natural profits as established firms and entrants alike can obtain them (p. 40).
[60] Gilbert (1989) considers such strategies are linked to "raising a rival's costs". As he puts it: "Demand- reducing strategies can be similar in effect to cost-increasing strategies."
[61] The historically costless acquisition of these rights does not withstand their asset character: In August 1997, EU and American government officials computed a possible merger between BA and AA to be "worth" up to 350 slots at Heathrow. In another case, after British Airways sold its stake in USAir in 1996, USAir sought to regain its slots to become once again a direct competitor of BA (USAir had given up its slots after its 1993 alliance with BA).
[62] See also the case of Alcoa, which was decided in 1945 (Scherer and Ross 1993, p. 454)
[63] Often such suppliers still remain within the group of the incumbent airline, thus controlling service rendered to third parties.
[64] The Air Letter, 20 May 1997.
[65] The Air Letter, 25 March 1998.
[66] 43% the Americas, 30% Southern, 23% Pacific and 4% Europe; BA Fact Book 1996, p. 61
[67] Frequent flyer programmes are a marketing strategy airlines use to encourage customer loyalty. Under these plans, passengers qualify for various awards by flying a specified number of miles with the sponsoring airline. The awards earned increase in attractiveness as the number of miles flown grows.
[68] Information source: Romios Voyage, Travel Agent, Geneva.
[69] In the EU smaller carriers tend to concentrate, at least initially, on predominantly business routes.
[70] Critics say that FFPs only exploit the principal-agent problem between employers and their employees travelling on business class.
[71] This is in contrast to the Bertrand-Nash concept of instantaneous price changes and quantity adjustments to clear markets.
[72] Neither are there uncertainty, switching costs or learning economies.
[73] In this specific context, cross-subsidies were considered absolute cost advantages. As such, they may form entry barriers and influence price sustainability.
[74] These nominal figures have not been deflated by any price index. It would have added only little insight, and finding a real price deflator common to all European countries might have proved very difficult indeed.
[75] The term sustainability is not without ambiguity: perfectly contestable markets mean low prices; whereas entry barriers may make high prices sustainable.
[76] After some exploratory clustering, we decided to use distance measures in favour of correlation coefficents: This produced 7 classes instead of 37 at a 5% significance level.
[77] We would also like to point out that standardized variables would have yielded 41 cases (on a 5% significance level). Moreover, variables that were privileged happen to be those representing output, something highly interesting when comparing price/output combinations within regression analysis.
[78] Closest in terms of Euclidean distance
[79] It is somewhat unfortunate that the clustering procedure chose this particular exemplar case to best represent Cluster 4. This is mainly due to the strong weighting attributed to capacity measures: The exemplar case shows a new city pair originating in London City Airport, not Heathrow, being served by Lufthansa with a Boeing 737. In this sense, the exemplar case stands for an incumbent who tries to copy entrants.
[80] These three factors explain cumulatively 69.85% of the total variance observed.
[81] Compare with Karnani & Wernerfelt, where incumbents choose either to: 1) do nothing, 2) defend its market, 3) counterattack, or 4) total attack.
[82] For comparison, we also ran the same regressions with a model involving a constant term. The model summary there yielded an R2 in a range between 20% and 25%, which we considered satisfactory.
[83] We suggest further research in this field, with a possible lead: CRS are particularly apt to leverage on multiple factors at the same time and to combine them differently in order to maximize revenues.
[84] To be more precise: airlines, whose stake in CRS were important relative to the size of their own route network, were the most advantaged.
[85] The airlines most often chose smaller aircraft for frequent flights, such as Lufthansa using Canadair and AvroJets instead of Boeing 737s. This was in line with our finding that the type of plane has an impact on prices.