: Airlines in Action Northeastern University & ISE KBTU EARIE, 2017
Incumbents and Entrants There are many studies of games between incumbents Analysis of games between incumbents and entrants is less rich and less helpful There are a few theoretical insights Most important may be that rational entry deterrence requires some irreversible action by incumbent This has focused research on capacity investment, not price But theory otherwise has offered little guidance Empirical work encounters difficulties Key is distinguishing strategic from normal business behavior Actual experience often seems to follow neither
Example of Issues: United at Newark Newark Airport has long been fortress hub for United Airlines High on list of most constrained airports in U.S. Substantial service to San Francisco, Los Angeles Virgin America sought to enter Newark to connect to its hubs on west coast Got 15 slots from bankruptcy of American Airlines in late 2012 In December 2012, Virgin announced it would start service to SFO, LAX in April 2013 In early April United announced immediate service increase Daily service to San Francisco increased from 7 to 16 departures Daily service to LAX from 6 to 14 departures Both United and Virgin persisted in capacity increases Prices fell dramatically Persisted through 2016 when Virgin acquired by Alaska
Questions, and Attempts at Answers Episode raises series of questions Why did United not take any pre-emptive actions? Given that, why did it take action at or after actual entry? Why did it persist in these actions? What was Virgin s strategy in persisting? This paper is effort to examine this and other experiences in order to better understand strategies Focus on airline industry Draws on existing literature
Some Useful Empirical Studies General studies of entry and entry deterrence Lieberman (1987) Dafny (2005) Ellison and Ellison (2011) Studies of entry and entry deterrence in airlines Goolsbee and Syverson (2008) Daraban and Fournier (2008) Kim (2009) Aguirregabiria and Ho (2010) Gayle and Wu (2013) Tan (2016)
Key Insight and Study Key to empirical work is to identify circumstances where entry deterring behavior would not be observed For example, when entry is already constrained, acts by incumbent unlikely to be motivated by entry deterrence Similarly, when entry deterrence infeasible Then, contrast incumbent actions with acts under other conditions Goolsbee and Syverson analyze response of various carriers to likely entry by Southwest Entry on route A to B defined as likely when Southwest enters airport B when it already serves airport A Find evidence of pre-emptive price decreases by carriers already serving A to B, but no capacity changes Interpret this as effort to build (invest in) customer loyalty before entry
Eight Propositions Regarding Entry, and Airlines I 1 Classic entry deterrence requires action with irreversible properties prior to entry 2 If incumbent acts only after entry, purpose must be something other than deterrence 3 Despite reversibility, price seems often to be used before entry, possibly as notification device 4 Pre-entry price decrease can also serve as investment in customer loyalty to handicap entrant
Eight Propositions Regarding Entry, and Airlines II 5 Capacity increases are in principle better suited as deterrent, but in airlines, capacity can be adjusted fairly quickly 6 Deterring low-cost carriers especially problematic for legacy airlines 7 Entry by legacy into route dominated by another legacy may be less frequent, and/or met with less resistance 8 Other factors may matter as well: hub vs. non-hub, concentration vs. unconcentrated route, incumbent s size and stake on route
This Study I Key is holding as much (nearly) constant as possible Focus on airline industry, one carrier at a time: United, US Air Same incumbent faces one entrant (at a time) into one market (hub or route) with known characteristics, e.g., Virgin at Newark Entry largely exogenously determined (result of slot divestiture) Evaluate incumbent responses against its actions in other markets that are not threatened by entry, as benchmark Familiar date base DB1B for tickets and prices T100 on departures and seats Standard screens and filters Standard list of control variables
Example 1: Virgin at Newark I United served 1462 routes throughout its system Of those, 93 routes in and out of Newark Virgin launched service on two of those routes in 2013q2 United responded Model compares United s response to entry by Virgin with its actions on Newark routes not subject to entry in same time period Response in terms of price, departures, seats Hold constant the incumbent, the entrant, the airport, the time period
Model of Incumbent Response to Entry y rt = β 0 + 5 β t (D t Airline Route r ) + γx rt + ɛ rt t= 6 Controls for HHI on route and at airports, distance, plus fixed effects for carrier-route and carrier-quarter-route Roughly similar to Goolsbee and Syverson
United Response to Virgin Entry on (EWR-SFO) I
United Response to Virgin Entry on (EWR-SFO) II ln(price) ln(departures) ln(seats) Virgin Route - not flying (t 0 6) 0.049* 0.263 0.275* (0.029) (0.168) (0.162) Virgin Route - not flying (t 0 5) -0.019 0.209 0.209 (0.024) (0.165) (0.160) Virgin Route - not flying (t 0 4) 0.031 0.567*** 0.624*** (0.021) (0.158) (0.157) Virgin Route - not flying (t 0 3) 0.174*** 0.744*** 0.732*** (0.022) (0.175) (0.170) Virgin Route - not flying (t 0 2) 0.113*** 0.576*** 0.492*** (0.027) (0.152) (0.144) Virgin Route - not flying (t 0 1) 0.225*** 0.497*** 0.468*** (0.027) (0.182) (0.181) Virgin Route - entered (t 0 ) -0.123*** 0.876*** 0.861*** (0.031) (0.154) (0.152) Virgin Route - flying (t 0 + 1) -0.173*** 1.593*** 1.541*** (0.032) (0.234) (0.226) Virgin Route - flying (t 0 + 2) -0.165*** 1.420*** 1.400*** (0.039) (0.201) (0.196) Virgin Route - flying (t 0 + 3) -0.220*** 1.021*** 0.988*** (0.032) (0.209) (0.206) Virgin Route - flying (t 0 + 4) -0.151*** 1.480*** 1.457*** (0.030) (0.203) (0.197) Virgin Route - flying (t 0 + 5) -0.068** 1.355*** 1.308*** (0.032) (0.187) (0.180) Carrier-Route FE Yes Yes Yes Carrier-Quarter FE Yes Yes Yes Year-Quarter FE Yes Yes Yes R 2 0.792 0.898 0.903 Observations 1329 1329 1329
Results for Virgin at Newark No indication that United changed pre-entry price Consistent with skepticism about price as deterrent Seemed reconciled to inevitable entry (due to slot acquisition) Slight increase in capacity before entry, but not conclusive Increase in departures and seats might thin out the market awaiting Virgin But then after entry, capacity increased dramatically, causing price to fall by 12-20 percent Could reflect notification, or investment in customer loyalty Suggests some commitment to maintaining dominance at EWR Same pattern for United on LAX route entered by Virgin
Example 2: Southwest at Newark Another exogenously determined entry: Condition of approval of United-Continental merger (2010) was slot divestiture to some LCC Southwest acquired slots at Newark Much bigger network, longer experience in US markets than Virgin Southwest started service from Newark in early 2011 To airports adjacent to two United hub cities Midway, but not O Hare in Chicago Hobby, but not Houston Intercontinental Also, to St. Louis, plus a couple other destinations
Results for Southwest at Newark First, on Newark-Chicago, United s actions: Price Departures Seats Before 0 0 0 After * 0 0 Much the same set of actions for Newark-Houston Implies United largely ignored Southwest entry into adjacent airport Could be due to limited competition from adjacent airport, or doubts that it could deter anyway Some indication of post-entry price reduction In Houston, slight indication of re-entry capacity increase
Example 3: Southwest to St. Louis I Southwest entered service to St. Louis in 2011q1 Route was not previously served by United, but it started up service shortly in 2011q2, right after Southwest United maintained service for 3 quarters, then withdrew Price Departures Seats After + +
Example 3: Southwest to St. Louis II
Implications Clearly not defending a route Nor deterring entry Nor building customer loyalty Possible explanations Establish reputation for toughness Defend Newark from progressive entry Notification strategy Similar to American s response to entry into DFW in 1994 SunJet opened from DFW to Long Beach in 1994 American did not serve, but launched competing service After SunJet withdrew, American ceased service DOJ sued for this and similar conduct by American, but lost in court
Second Setting: US Air on Boston-Philadelphia As prelude to constructing larger data base, we examine one additional carrier on an important specific route US Air has dominated Boston-Philadelphia route for 20 years Important and profitable route Dominance has not gone unchallenged: Delta entered in 2001q1, but exited in 2002q4 AirTran entered in 2003.2, but exited in 2008 Southwest started service from two airports near Boston in 2004q2 Then entered BOS-PHL itself in 2010 But exited all of them entirely in 2012 JetBlue entered in 2013q2 Series of attempts at entry offers insights into Legacy carrier entry Nearby airport entry Repeated entry
Example 4: Delta Entry Price Departures Seats Before 0 0 0 After 0 0 0 Implies absence of any response by US Air, before or after Could be that non-lcc entrant poses less of a threat Consistent with other evidence about diminished competition between legacies
Example 5: AirTran Entry Price Departures Seats Before 0 0 0 After + + Implies that LCC entrant does trigger response But only after entry no expectation of actual deterrence Both capacity and price changes Again, sending notification Also, reducing immediate profit from entering
Example 6: Southwest Entry into Nearby Airports Southwest entered service from Philadelphia to Providence RI and Manchester NH Each about 50 miles from Boston s airport Analyze US Air s actions on BOS-PHL in response US Air s response to both entries similar: Price Departures Seats Before 0 0 After 0 0
Implications US Air perceived significant threat from Southwest, even if only at nearby airports Use of price could not be intended to deter Rather, could have signaled displeasure Or warning not to open up at Boston s airport Regardless, made entry less immediately profitable for Southwest Interestingly, no capacity response by US Air
Example 7: Southwest Entry into BOS-PHL Price Departures Seats Before + 0 0 After 0 0 When Southwest started service at Boston s airport, no indication of deterrence or punishment strategy from US Air Post-entry price was lower but only relative to unusually high pre-entry price No change in pre or post entry capacity US Air appears reconciled to successful entry Maybe tried to extract profits before inevitable actual entry, but then forced to reduce price
Example 8: JetBlue Entry Price Departures Seats Before + 0 0 After 0 0 US Air s actions much like those with respect to Southwest in BOS-PHL Higher prices before, lower upon actual entry No changes in capacity Suggests US Air doubted it could actually deter Again, price pattern suggests pre-entry harvesting but post-entry determination to remain in market
Conclusions... Neither carrier seems to have complete and systematic strategy to deter Seem to respond in case-specific ways No clear differences between United and US Air Price seems to be often used as strategy And often only after actual entry Much less often used are departures and seats Responses to LCC vs. legacy entry seems to differ Even though legacy carrier would seem more vulnerable Even at nearby airports, Southwest prompts vigorous response Notification strategy would seem to be in evidence
... and Questions Need to further test entry by legacy Need to further test notification strategy vs. loyalty building Need to examine reluctance to use capacity but common use of price Need to find cases of dominant LCC responding to entry Need to examine wider range of experiences This paper is an exercise in formulating better hypotheses