3. Airline Choices Under Noise Constraints
3.1. The case of a cumulative constraint
Suppose that noise regulation takes the form of a cumulative noise constraint, which imposes
a common limit on the total noise generated by each airline, being written nifi ≤ L,
i = 1, 2. Since this constraint will bind in equilibrium, it can be used to eliminate n1, which
can be written n1 = L/f1. Substituting in (8), the multiplicative p1 − τ − /n1 term becomes
p1 − τ − f1/L. The resulting objective function then depends only on the choice variables p1
and f1.
In modeling airline decisions, an important question concerns the timing of choices. While
both flight frequencies and fares can be adjusted relatively easily, fares appear to be more
flexible than frequencies. As a result, under the most realistic approach, fares would be chosen
conditional on frequencies, which themselves would be set in a first stage, taking the secondstage
impact on fares into account (the outcome is a subgame perfect Nash equilibrium).
The analysis eventually adopts this approach, but it is useful to start by considering the less-
8
realistic case where fares and frequencies are chosen simultaneously in Nash fashion. Under this
approach, airline 1 simultaneously sets the levels of p1 and f1, taking p2 and f2 as parametric.
The first-order conditions for the simultaneous-choice problem are
∂π1
∂p1
= q1 − p1 − τ − f1/L
α
= 0 (9)
∂π1
∂f1
= −
L
1
2
− 1
α
[p1 − p2 + γ/f1 − γ/f2]
+
γ(p1 − τ − f1/L)
αf2
1
− θ = 0. (10)
While the derivatives ∂2π1/∂p21
and ∂2π1/∂f2
1 are easily seen to be negative, the remaining
second-order condition (positivity of the Hessian matrix of π1) is assumed to hold.11
Given the symmetry of the setup, the equilibrium fares and frequencies will be symmetric
across carriers, with the values denoted p and f, and each airline’s traffic will equal 1/2.
Imposing the latter condition in (9) and rearranging, the fare can be written
p = α/2 + τ + f/L. (11)
Thus, the fare equals the full marginal cost of a seat, given by the operating cost τ plus noise
abatement cost per seat (f/L),12 plus a markup that depends on brand loyalty. Note that,
as the extent of brand loyalty rises, leading to a higher α/2, a carrier loses less traffic from
raising its fare, so that p increases.
After substituting (11) into (10) and imposing symmetry, the equation can be solved for
the equilibrium frequency, which equals
f =
γ
2θ + /L
. (12)
This solution shows that frequency is increasing in the disutility of schedule delay (γ) and
decreasing in the fixed cost per flight (θ), both natural conclusions. Since the fare in (11)
increases with frequency, p is then increasing in γ and decreasing in θ. Note that the last
9
conclusion, which says that p is inversely related to the fixed cost per flight, is counterintuitive.
The explanation is that the reduction in service quality caused by the higher θ reduces
willingness to pay for air travel. Observe also that f is independent of the marginal seat cost
(τ ) and the extent of brand loyalty (α).13
The key comparative-static question, however, concerns the effect of the noise limit L on
the airline’s choice variables. When the noise constraint is relaxed by increasing L, (12) shows
that frequency rises, with the denominator expression becoming smaller. This conclusion is
natural since, as L rises, the airline can raise f for a given n without violating the constraint.
With L and appearing in ratio form in (12), the opposite conclusion applies to an increase
in : by raising the cost per flight, a higher noise-abatement cost lowers frequency.
Turning to the impact of L on p, the effect is not immediately apparent given that p has an
direct inverse dependence on L from (11) as well as an indirect dependence via f. However, after
substituting (12) into (11) and rearranging, the last term reduces to
γ2/(2θL2 + L), which
is decreasing in L. Therefore, relaxation of the noise constraint leads to a lower equilibrium
fare, a conclusion that shows how the cost of airline travel is affected along with service quality
(f) by the stringency of noise regulation. Since L and enter asymmetrically in the previous
expression, their impacts are no longer mirror images. But it is easily seen that the expression
is increasing in , so that the fare naturally rises with the cost of abatement.
Aircraft quietness and size, n and s, can be recovered from the f solution, allowing further
comparative-static effects to be derived. Since s = q/f = 1/2f, the parameter impacts on
aircraft size are opposite to those on flight frequency. With n = L/f, substitution of (12)
yields
n =
2θL2 + L
γ
, (13)
so that n is increasing in L and , both natural conclusions. Thus, aircraft quietness falls (n
rises) as the noise constraint is relaxed or the cost of abatement increases. In addition, (13)
shows that a higher fixed cost per flight raises n, while an increase in the delay disutility has
the opposite effect. All of these comparative-static results are summarized in Table 1.