Preliminary and Incomplete—Comments Welcome Incentives and Competition in the Airline Industry Rajesh K. Aggarwal D’Amore-McKim School of Business Northeastern University Hayden Hall 413 Boston, MA 02115 [email protected]Carola Schenone McIntire School of Commerce University of Virginia Charlottesville, VA 22904 [email protected]Draft: September 7, 2015
26
Embed
Preliminary and Incomplete—Comments Welcome › sites › default › files... · Preliminary and Incomplete—Comments Welcome Incentives and Competition in the Airline Industry
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Preliminary and Incomplete—Comments Welcome
Incentives and Competition in the Airline Industry
Rajesh K. Aggarwal D’Amore-McKim School of Business
Where is a route-carrier fixed effect; is a year-quarter fixed effect; and is an idiosyncratic
unobservable.
For financial performance data, the unit of observation is a carrier j, on a year-quarter t. The
econometric specification is,
_∈ , , , , , ,
where is a carrier fixed effect, is a year-quarter fixed effect, and an idiosyncratic unobservable.
10
4. Results
Table 3, Panel A, provides results from our baseline specification. The left three columns
examine the reduction in the minutes of delay. Airlines that institute an on-time arrival incentive
significantly decrease their arrival delays. A typical flight is approximately 135 minutes long from
Table 2, Panel B. Instituting an on-time arrival incentive reduces arrival delays by 11.1% of 135
minutes, or 15 minutes. For competitors, if a rival airline in the same market (routes or city-pairs)
initiates an on-time arrival incentive, then the competitor airline that did not introduce the incentive still
significantly reduces its arrival delays by 6.8% of 135 minutes, or 9.2 minutes. If both airline
competitors have an on-time arrival incentive, then there is no incremental reduction in delays. In this
case, both having an incentive offsets the reduction in delays from the rival having an incentive by an
incremental and statistically significant 6.9%. The net effect for both airlines is to reduce their delays
by 11% (11.1% + 6.8% - 6.9%) or 14.9 minutes.
The middle three columns restrict attention to situations in which the arrival delay is non-
negative. Here we exclude cases in which flights arrive early (prior to their scheduled arrivals). Flights
arriving early were included in our previous results, partially explaining the magnitude of the reduction
of the delays. When we exclude early arrivals, we find that instituting an on-time arrival incentive
reduces delays by 4.9%, or 6.6 minutes on a 135 minute flight. The effect on rivals is a reduction in
delays of 4.2%, or 5.7 minutes. If both competitors have on-time arrival incentives, then the total effect
is 4.9% + 4.2% - 3.8% = 5.3%, or a reduction in delays of 7.2 minutes. All of these effects are
statistically significant.
The right three columns consider the fraction of flights delayed more than 15 minutes, an
industry standard definition for a flight to not be on-time. Instituting an on-time arrival incentive
reduces the fraction of flights delayed more than 15 minutes by a significant 7.2%. For rivals, the
fraction of flights delayed more than 15 minutes is reduced by a significant 6.3%. If both competitors
11
on a route have on-time arrival incentives, then the effect of the reduction for both is attenuated by a
significant 5.0%, so that the total fraction of flights delayed more than 15 minutes is 7.2% + 6.3% -
5.0% = 8.5% for both airlines.1
We next dig deeper into the sources of the reduction in the arrival delays. In Table 3, Panel B,
we consider departure delays. In the left three columns, for airlines instituting an on-time arrival
incentive, their departure delays decrease by a significant 5.9% or 8.0 minutes. For rival airlines on the
same route without an incentive, their departure delays decrease by a significant 1.7% or 2.3 minutes.
When both have an incentive, these effects are significantly attenuated by 4.3%, so that the total
departure delay decrease for both competitors is 5.9% + 1.7% - 4.3% = 3.3%, or 4.5 minutes. We see
similar effects for the non-negative departure delays (eliminating those flights that depart the gate early)
and for the fraction whose departures are delayed more than 15 minutes. For non-negative departure
delays when both have an incentive, the total decrease in departure delays is 4.1% + 2.6% - 3.8% =
2.9%, or 3.9 minutes. This 3.9 minutes reduction in departure delays accounts for 54% of the 7.2 minutes
reduction in arrival delays.
An interesting effect here is that both airlines instituting an incentive actually results in less of a
reduction in departure delays than does only one airline instituting an incentive. In principle, if one
airline introduces an incentive, the rival airline would be better off not introducing an incentive, while
still getting the benefit of a reduction in departure delays. Of course, reducing departure delays is not a
goal in and of itself, since the incentives are for on-time arrivals. The key point here is that airlines
seem to respond to an on-time arrival incentive by attempting to get flights to depart the gate closer to
on-time.
1 As a robustness check, we consider whether the reduction in arrival delays that we document in Table 3, Panel A, is concentrated in problem markets. We define problem markets as routes in which flights are delayed more than 15 minutes over 20% of the time prior to the initiation of the on-time arrival incentive. In unreported results, we find that essentially all of the reductions in delays happen in the problem markets.
12
Another possibility is that airlines might try to game the on-time arrival incentive. When faced
with a new incentive, executives might strategically increase the scheduled amount of time for a flight.
To examine this, we look at both the scheduled amount of time for flights and the actual duration of the
flights in Table 4. The left panel shows that an airline instituting an on-time arrival incentive increases
its scheduled flight times by 0.5%, or about 41 seconds. While statistically significant, this effect is not
economically meaningful. A rival airline will increase its scheduled flight times by 0.9%, or about 1.2
minutes. If both have an on-time arrival incentive, then both will incrementally increase their scheduled
flight times by 1.8% or 2.4 minutes, so that the total effect would be 0.5% + 0.9% + 1.8% = 3.2%, or
4.3 minutes, suggesting some degree of collusive behavior.
The middle panel considers actual elapsed times. An airline instituting an on-time arrival
incentive sees its actual elapsed times increase 0.4% or about 32 seconds. A rival airline decreases its
elapsed flight times by 0.4%, or about 32 seconds. While statistically significant, these effects are not
economically meaningful. If both have an on-time arrival incentive, then both will incrementally
increase their elapsed flight times by 0.9% or 1.2 minutes, which is the same as the total effect (0.4% -
0.4% + 0.9%). This is again a relatively small magnitude, but suggests that airlines increase their
scheduled times by much more than the actual times needed to fly these routes. In situations in which
both airlines on a route have an incentive, scheduled times increase by 4.3 minutes while actual elapsed
times only increase by 1.2 minutes. This incremental 3.1 minutes of scheduled time can account for
43% of the reduction of 7.2 minutes in non-negative arrival delays when both competitors have an on-
time arrival incentive.
The right panel considers markets served by the airlines in response to the on-time arrival
incentive. An airline that institutes an on-time arrival incentive expands the number of routes that it
serves by a significant 4.0%. This suggests that the reduction in delays is not due to the creation of
more slack by eliminating routes. The rival airline by contrast reduces the number of routes served by
13
a significant 4.8%, suggesting some competition in strategic substitutes. If both have an incentive, the
routes served decrease by 1.8% (4.0% - 4.8% - 1.0%).
Table 5 digs deeper into capacity. Panel A considers seats, passengers, and quarterly load
factors. Instituting an on-time arrival incentive is associated with an increase in seats (3%) and
passengers (5%) on each route. For the rival airline on the same route, seats shrink by 4% and
passengers shrink by 3%, suggesting a significant increase in competition on these routes that the rival
then accommodates. If both have an incentive, seats and passengers both shrink by an incremental 8%.
Not surprisingly, the load factor increases for the incentive-initiating airline. It also increases for the
rival airline, as the number of seats shrinks more than do passengers. Load factor does not change if
both have an incentive, as seats and passengers shrink by similar amounts.2
Panel B considers departures scheduled and departures performed. An incentive-initiating
airline increases both its departures scheduled and performed by 7%. The rival airline on the same
routes will increase its departures scheduled by 2%, but with no significant difference in departures
performed. If both have an incentive, then the incremental effect on departures scheduled and
performed is a decrease of 11%, so that the number of departures scheduled and performed shrinks by
2% and 4%, respectively, if both have an incentive.
Table 6 considers the effect of on-time arrival incentives on financial performance. Here our
sample consists of airline-year observations rather than route-airline-quarter observations. Given that
actions are taken to improve on-time arrivals once an incentive is initiated, we ask whether this then
translates into improved financial performance. For an incentive-initiating airline, we find that revenues
and expenses as a fraction of assets both increase (with marginal significance), but there is no significant
change in profits and the magnitude of the coefficient on profits is economically very small. For the
rival airline, all of the coefficients (on revenues, expenses, and profits) are insignificant. If both have
2 Load factors take into account passenger seat miles, so it is possible that airlines may optimize their routes to focus on longer haul routes in response to an on-time arrival incentive. This may be optimal if it is easier to have longer haul flights arrive on-time as departure delays can be made up while flying if the distance is greater.
14
an incentive, there is a marginally significant incremental reduction in revenues and expenses, but again
no significant effect on profits. Thus, we do not see significant effects on financial performance from
the introduction of on-time arrival incentives.
Our previous results use pooled data and specifications that are saturated with fixed effects to
absorb variation by quarters, markets, and carriers. As an alternative, we can use the richness of our
data to examine changes in on-time performance around the introduction of an incentive. We consider
a difference-in-difference specification around the introduction of an on-time arrival incentive. Each
carrier introduces an on-time arrival incentive at a different point in time (one carrier in our sample,
JetBlue, does not merge with another carrier and does not introduce an incentive over our sample
period). We use this fact to estimate the reduction in arrival delays for a carrier around the introduction
of the incentive relative to carriers that do not have an incentive at that time. The advantage of this
approach is that it allows us to estimate reduction in arrival delays by carrier at the time of the
introduction of the incentive. The disadvantage of this approach is that we do not control for route-
carrier fixed effects.
For each of the six carriers that instituted an incentive, we select a time window of one year
before and after the carrier introduced an incentive. We then select as control carrier those carriers that
have not had an incentive introduced at any time before the selected time window. For example, Delta
introduced an incentive in 1997, the only carrier that had an incentive in place before 1997 is
Continental, so we include as control groups for Delta all carriers, excluding Continental. For carriers
that instituted an incentive in the later part of our full sample period, such as Southwest, the set of
control carriers is limited to those carriers that have not introduced an incentive at any point in time
within our sample years of 1993-2013, and they include America West, Northwest, and Jet Blue. We
also deal with carriers that introduced an incentive in overlapping time windows as follows: US Airways
introduced an incentive in 2008 and Southwest in 2009, therefore in the one year time window we
15
exclude Southwest as a control for US Airways’ treatment. We run a difference-in-difference
specification of the type:
_ __ _ _ _ _ _ ∗ _
where _ is a categorical variable, equal to 1 if the operating carrier is carrier i;
_ _ is a categorical variable equal to 1 if the observation corresponds to the time
window for the year(s) after the carrier implemented an incentive.
Table 7 contains the results. The coefficient of interest is _ , the coefficient on the
introduction of the incentive interacted with the carrier. Continental, Delta, and US Airways, show
significant reductions in delays after the introduction of an incentive. For example, the coefficient for
Continental of -1.90 implies that Continental reduces its arrival delays by 1.9 minutes in from the year
after the introduction of the incentive relative to the year prior to the incentive and relative to rival
carriers without an incentive (in this case, all of them, as Continental was the first to introduce an
incentive). Interestingly, United Airlines shows an increase in delays after the introduction of an on-
time arrival incentive relative to rival carriers. We note that United’s introduction of an on-time arrival
incentive coincided with its emergence from bankruptcy. For the other two carriers, American and
Southwest, there is no significant effect on delays.
5. Conclusion
We find that airlines reduce the length and frequency of arrival delays in response to the
introduction of on-time arrival bonuses for their executives. Because these bonuses were staggered in
their introduction across airlines, we are able to identify the effect of the bonus on arrival delays.
Because on-time arrival is an important quality metric for passengers and airlines, we can directly
16
observe the impact of an incentive on the outcome variable it is meant to influence. Interestingly, we
find that competitors increase their on-time arrival performance in response to a rival introducing an
incentive, even if the competitor does not have an on-time arrival incentive in place. This suggests that
competition in quality is competition in strategic complements. Consistent with this, we find that there
is no improvement in airline profitability after the introduction of incentives. Instead, all of the
improvement in on-time performance seems to go to consumer surplus.
We also find some evidence that part of the improvement in on-time performance comes from
strategic gaming by airlines. Especially when two airlines on a route have on-time performance
incentives, airlines will increase the scheduled time of flights. This effect can explain 43% of the
improvement in on-time performance when two competitors each have an incentive. Another 54% of
the improvement in on-time performance when two competitors each have an incentive can be attributed
to a reduction in departure delays (getting airplanes out of the gate on time), and this represents most of
the positive effect of incentives. When only one airline introduces an on-time arrival incentive, then
the effect of strategic gaming is negligible, and most of the improvement comes from the reduction in
departure delays. These results suggest that the introduction of on-time arrival incentives by multiple
carriers in a market may in fact foster tacit collusion to the benefit of managers.
17
References
Aggarwal, Rajesh K., and Andrew A. Samwick (1999), ““Executive Compensation, Relative Performance Evaluation, and Strategic Competition: Theory and Evidence,” Journal of Finance 54: 1999-2043.
Ciliberto, Federico, and Carola Schenone (2012a), “Are the Bankrupt Skies the Friendliest?” Journal of Corporate Finance, 18: 1217-1231.
Ciliberto, Federico, and Carola Schenone (2012b), “Bankruptcy and Product-Market Competition: Evidence from the Airline Industry,” International Journal of Industrial Organization, 30: 564–577.
Holmstrom, Bengt and Paul Milgrom (1987), “Aggregation and Linearity in the Provision of Intertemporal Incentives,” Econometrica, 55: 303-328.
Table 1: Summary Statistics for On Itself, On Others, and On Both The data for incentives is collected from firm proxy statements. The sample of carriers and flights is from the On‐Time Performance Database in Panel A, and from the T100 Dataset in Panel B; Schedules B1 and P1.2 of the Domestic Segment Form 41‐Financial Data. The Bureau of Transportation Statistics (BTS) provides all three datasets. The sample ranges from January 1993 to December 2011. For panels A and B, the entries in this table correspond to the market‐carrier‐year‐quarter count for the categorical variables On Itself and On Others. On Itself is defined as _ where _ equals 1 if carrier c has an
incentive at time t and operates in route m and the observation is for carrier c. On Others is defined as
_ where _ equals 1 if carrier c has an incentive at time t, and carrier c and j
operate route r at time t, and the observation corresponds to carrier j. On Both equals 1 when 1. Thus On Itself captures the effect of the on‐time arrival incentive for the carrier receiving the incentive;
On Itself captures the effect of the on‐time arrival incentive for competitors of the carrier receiving the incentive; and On Both captures the effect of multiple carriers having an on‐time incentive at the (overlapping) same time and operating in the same route. Panel C reports financial data and uses the 2 times lagged variables On Itself and On Others.
Panel A: On Itself, On Others and On Both – On Time performance Note that there are 15,652 observations for which more than one carrier has an incentive at (partially) overlapping times and operate in the same market at that time. Thus, in our estimation sample, On Both equals one for 15,652 year‐quarter‐market‐carrier observations.
On Itself On Others 0 1
Totals
0 105,606 72,146 177,7521 13,244 15,652 28,896
118,850 87,798 206,648
Panel B: On Itself, On Others and On Both – Capacity T100 Dataset The entries in this table correspond to the market‐carrier‐year‐quarter count for the categorical variables On Itself and On Others for the capacity sample built using from the T100 Domestic Segment of Form 41: Air Carrier Statistics‐US Carriers, reported by the BTS. Stats on data for estimation sample.
On Itself
On Others 0 1
Totals
0 151,443 112,598 264,0411 23,270 28,631 51,901
174,713 141,229 315,942
Panel C: Lag x2 On Itself, Lag x2 On Others and Lag x2 On Both – Financial Data The entries in this table correspond to the market‐carrier‐year‐quarter count for the categorical variables Lag x2 On Itself and Lag x2 On Others for the financial data sample sourced from the BTS under Schedules B1 and P1.2 of the Domestic Segment Form 41‐Financial Data. Note that there are 174 observations for which more than one carrier has an incentive at (partially) overlapping times and operate in the same market at that time.
Lag x2 On Itself
Lag x2 On Others 0 1
Totals
0 56 13 691 353 168 521
409 181 590
Table 2: Summary statistics. Arrival and departure delays (Panel A), and computer reservation system (CRS) scheduled flight times and actual flight times, as well as percent of markets served (Panel B). Panel C reports summary statistics for carrier capacity: Scheduled and Performed flights, seats and passengers per seats. Finally, Panel D reports statistics for financial data: The ratio of operating revenues expenses and profits to total assets. Data source: On‐Time Performance Database provided by the BTS, for Panels A and B; the T100‐Domestic Segment of Form 41: Air carrier Statistics‐US Carriers, reported by the BTS, Panel C; and from the BTS under Schedules B1 and P1.2 of the Domestic Segment Form 41‐Financial Data, Panel D. The sample ranges from January 1993 to December 2011. On Itself is defined as _ where _
equals 1 if carrier c has an incentive at time t and operates in route m and the observation is for carrier c. On Others is defined as _ where _ equals 1 if carrier c has an incentive at time t, and carrier c
and j operate route r at time t, and the observation corresponds to carrier j. On Both equals 1 when 1. Thus On Itself captures the effect of the on‐time arrival incentive for the carrier receiving the incentive; On Others
captures the effect of the on‐time arrival incentive for competitors of the carrier receiving the incentive; and On Both captures the effect of multiple carriers having an on‐time incentive at the (overlapping) same time and operating in the same route.
Panel B: CRS and actual elapsed time and percent of markets served
CRS Elapsed Time Actual Elapsed time Pct Markets Served
On Itself = 0 On Others = 0 N=105,606
133.042.20
133.05(2.20)
15.64(0.01)
On Itself = 1 On Others = 0 N=72,146
143.537.51
143.39(7.50)
10.83(0.00)
On Itself = 0 On Others = 1 N=13,244
131.269.82
129.96(9.80)
15.97(0.49)
On Both =1 N=15,652
109.1911.16
111.52(11.06)
14.48(0.86)
Table 2 (Continuation)
Panel C: Capacity
Departures Scheduled Departures Performed Seats Quarterly Load Factor
On Itself = 0 On Others = 0 Obs=151, 443
329.87 (244.60)
323.01(238.18)
42,384 (34008)
0.66(0.14)
On Itself = 1 On Others = 0 Obs=112,598
312.37 227.617
299.91221.39
33,308 (32349)
0.73(0.13)
On Itself = 0 On Others = 1 Obs=23,270
359.83 (266.50)
350.15(256.99)
44,116 (35,422)
0.71(0.13)
On Both =1 Obs=28,631
322.17 (234.09)
310.74(228.38)
39,879 (34,433)
0.77(0.11)
Panel D: Financial Performance
Operating Revenues to Total Assets
Operating Expenses to Total Assets
Operating Profits
Lag x2 On Itself = 0 Lag x2 On Others = 0 56 Obs
0.20(0.05)
0.190.05
0.01(0.02)
Lag x2 On Itself = 1 Lag x2 On Others = 0 13 Obs
0.25(0.03)
0.24(0.03)
0.02(0.01)
Lag x2 On Itself = 0 Lag x2 On Others = 1 353 Obs
0.18(0.07)
0.17(0.07)
0.01(0.02)
Lag x2 On Both =1 168 Obs
0.16(0.05)
0.16(0.05)
‐0.004(0.02)
Table 3: The Impact of managerial incentives for on‐time arrivals on delays at arrival (Panel A) and departures (Panel B).The regression is of the form
_ _∈ , , , , , ,
is one of three measures of arrival or departure delays (minutes delayed, non‐negative minutes delays, fraction of flights delayed more than 15 minutes), for carrier j in
route r at year‐quarter t. Minutes of delay for j in route r at t, is the average, across the daily delays in that route and time. equals 1 when carrier j at time t has an
incentive and the observation corresponds to carrier j at t; is 1 when carrier j at time t serving route r, has an incentive and the observation corresponds to a
competitor of j in route r at t. equals 1 when equal 1. _ controls for periods in which carrier c serving route r
operates under bankruptcy protection; _ controls for the presence of carrier c in route r, at time t. is a route‐carrier fixed effect; is a year‐quarter fixed effect;
and is an idiosyncratic unobservable. Data source: On‐Time Performance Database provided by the BTS.
Incentive On Itself On Others On Both On Itself On Others On Both On Itself On Others On Both
Percent Departure delay ‐0.059 (0.004)
‐0.017(0.007)
0.043(0.008)
‐0.041(0.003)
‐0.026(0.004)
0.038(0.005)
‐0.052(0.003)
‐0.046(0.005)
0.061(0.006)
Observations 206,648 206,648 206,648
Groups 8,051 8,051 8,051
Min obs. per group 1 1 1
Avg. obs. per group 26 26 26
Max obs. per group 84 84 84
F‐Stat 324 545 485
R‐2 within 0.1341 0.2067 0.1885
Bankruptcy Controls Yes Yes Yes
Carrier Active in Market Yes Yes Yes
Table 4: The Impact of managerial incentives for on‐time arrivals on actual flight times and on the airline’s scheduled flight time as reported to the Computer Reservation System (CRS) The regression is of the form
_ _∈ , , , , , ,
is one of three variables: CRS_Timerjt carrier j’s flight time for route r at time t as reported by j in the Computer Reservation System (CRS); Actual_Elapsed_Timerjt is
the actual flight time recorder for carrier j in route r at time t; and Pct_Routes_Servedrjt. is the ratio of the number of routes carrier j served at time t to the number of routes all carriers served at time t. equals 1 when carrier j at time t has an incentive and the observation corresponds to carrier j at t; is 1 when
carrier j at time t serving route r, has an incentive and the observation corresponds to a competitor of j in route r at t. equals 1 when
equal 1. _ controls for periods in which carrier c serving route r operates under bankruptcy protection;
_ controls for the presence of carrier c in route r, at time t. is a route‐carrier fixed effect; is a year‐quarter fixed effect; and is an idiosyncratic unobservable. Data source: On‐Time Performance Database provided by the BTS.
CRS Time Actual Elapsed Time Log (Percent Routes Served)
Incentive
On Itself On Others Both On Itself On Others Both On Itself On Others Both
Percent Departure delay 0.005 (0.001)
0.009(0.002)
0.018(0.002)
0.004(0.001)
‐0.004 (0.002)
0.009(0.002)
0.04(0.002)
‐0.048(0.003)
‐0.01(0.003)
Observations 206,648 206,648 206,648
Groups 8,051 8,051 8,051
Min obs. per group 1 1 1
Avg. obs. per group 26 26 26
Max obs. per group 84 84 84
F‐Stat 166 151 177
R‐2 within 0.0737 0.0673 0.0780
Bankruptcy Controls Yes Yes Yes
Carrier Active in Market Yes Yes Yes
Table 5: The Impact of managerial incentives for on‐time arrivals on a carrier’s capacity and utilization The regression is of the form
_ _∈ , , , , , ,
is one of three variables in Panel A: Seatsjrt is the number of available seats carrier j offers in route r at time t, and is calculated as the sum of the daily number of
seats available in the airplanes used by j in r during year‐quarter t. Passengersjrt is the sum of all passengers carrier j transported in route r at time t. Pass_to_Seatsjrt is the sum of the daily ratios of passengers transported over seats available. In Panel B: is the total number of departures scheduled Departures_Scheduledjrt, and
performed Departures_Performedjrt, by carrier j in route r at time t. equals 1 when carrier j at time t has an incentive and the observation corresponds to
carrier j at t; is 1 when carrier j at time t serving route r, has an incentive and the observation corresponds to a competitor of j in route r at t.
equals 1 when equal 1. _ controls for periods in which carrier c serving route r operates under bankruptcy
protection; _ controls for the presence of carrier c in route r, at time t. is a route‐carrier fixed effect; is a year‐quarter fixed effect; and is an
idiosyncratic unobservable. Data Source: T100‐Domestic Segment of Form 41: Air carrier Statistics‐US Carriers, reported by the BTS
Table 6: The Impact of managerial incentives for on‐time arrivals on carrier profitabilityThe regression is of the form
_ _ _ _ _ _
_∈ , , , , , ,
is one of three measures of financial performance:
,
,
for carrier j in year‐
quarter t. _ controls for periods in which carrier c operates under bankruptcy protection; is a year‐quarter fixed
effect; is a carrier fixed effect, and is an idiosyncratic unobservable. Data Source: BTS under Schedules B1 and P1.2 of the Domestic Segment Form 41‐Financial Data.
Incentive
Lag x2 On Itself
Lag x2 On Others
Lag x2 Both
Lag x2 On Itself
Lag x2 On Others
Lag x2 Both
Lag x2 On Itself
Lag x2 On Others
Lag x2 Both
Percent change 0.05 (0.03)
‐0.02 (0.03)
‐0.05(0.03)
0.05(0.03)
‐0.004 (0.03)
‐0.05(0.03)
0.003 (0.011)
‐0.010(0.011)
0.001 (0.011)
Observations 590 590 590
Groups 9 9 9
Max obs. per group 40 40 40
Avg. obs. per group 66 66 66
Max obs. per group 74 74 74
F‐Stat 4.68 2.17 10.84
R‐2 within 0.4097 0.2433 0.6529
Op. carrier FE Yes Yes Yes
Year Quarter FE Yes Yes Yes
Table 7: Difference in Difference
For each of the six carriers that received an incentive, we select a time window of one year before and after the carrier introduced an incentive. We then select as control
carrier those carriers that have not had an incentive introduced at any time before the selecteded time window. For example, Delta introduced an incentive in 1997, the
only carrier that had an incentive in place before 1997 is Continental, so we include as control groups for Delta all carriers, excluding Continental. For carriers that
instituted an incentive in the later part of our full sample period, such as Southwest, the set of control carriers is limited to those carriers that have not introduced an
incentive at any point in time within our sample years of 1993‐2013, and they include America West, Northwest, and Jet Blue. We run a diff‐in‐diff specification of the
type:
_ _ _ _ _ _ _ _ ∗ _
where _ is a categorical variable, equal to 1 if the operating carrier is carrier i; _ _ is a categorical variable equal to 1 if the observation
corresponds to the time window for the year after the carrier implemented an incentive. The coefficient of interest is _ .