Self-Financing Theorem (The General Case)

1. Optimal toll revenue exactly covers the optimal capacity cost when production of transport services is (local) constant-return-to-scale. 2. Long-run total cost of the transport system T C (f ) is such that T C (f ) = min f c (f, w) + k (w)
w

where f is ow volume, w is capacity, c is user cost, and k is the capacity cost. 3. When T C (f ) exhibits (local) constant return to scale, we have AC = M C. 4. Optimal tolling implies that toll equals to M EC, and price equals to M C. 5. Together we have AC = M C = P where AC MC P = = = T C (f ) k (w) =c+ f f c c+f f c+

with being the optimal toll, which gives k (w) = f .

Self-Financing Theorem (a Special Case of Auto Transport)

1. Optimal toll revenue exactly covers the optimal capacity cost when (i) user cost is homogeneous of degree zero in volume and capacity and (ii) capacity construction exhibits (local) CRS. 2. User cost c (f, w) is homogeneous of degree zero if c (f, w) = c (tf, tw) t. 3. Then aggregate user cost f c (f, w) is homogeneous of degree one. 4. Therefore, if k (w) is also homog. of 1 , so is the entire f c (f, w) + k (w). 5. Let T C (f, w) = f c (f, w) + k (w). Then by Eulers theorem we have that where cf = c/f and cw = c/w. That is, f c (f, w) + k (w) = T C (f, w) = f T C f + wT C w

f [c + f cf ] + w [f cw + k ] .

6. But here, if capacity w is optimal, f cw + k = 0 by FOC. 7. Thus we have k (w) = f [f cf ] where f cf = M EC = under optimal tolling.

Comments on Airport Capacity and Congestion When Carriers Have Market Power by Anming Zhang and Yimin Zhang
For the International Symposium on Spatial Economics and Transportation June 13th, 2005
By Yuichiro Yoshida
National Graduate Research Institute for Policy Studies

Overview of the Paper

The paper pioneers in investigating the effect of carriers market power on airport capacity and financial viability Three airport types (welfare-maximizing, profitmaximizing, and budget-constrained public airports) are examined for different market structures Relevant results of the paper includes that when air carriers have market power, they internalize congestion externality and thus deprives the welfare-maximizing airport authority of the ability to raise revenue necessary for the optimal capacity construction

Structure of the Analysis

Stage 1 Airport , K Stage 2 Carriers ... Maximizes SW w/ or w/o constraint, or maximizes profits: , K = arg max SW (s.t. Q - c0 Q - rK = 0) or, = arg max " = Q - c0 Q - rK ... Maximizes profits given others quantity as well as and K:

qi => Q
Stage 3 Passengers

qi = arg max "i = qi (! - D) - ciqi - iqi

... Full price is determined through the inverse demand function:

! = ! (Q)

Summary of Results
Social Welfare Maximization Competitive > Capacity is optimal. > Airport charge equals the social marginal cost. > Ticket price paid by passenger is equal to social marginal cost > For symmetric Cournot oligopoly, capacity is optimal. > Carriers partially internalizes the externality. > Airport charge is less than the social marginal cost. > Ticket price paid by passenger is equal to social marginal cost > Capacity is optimal. > Carrier fully internalizes the externality. > airport charge is much less than the social marginal cost, could be negative. > Ticket price paid by passenger is equal to social marginal cost Profit Maximization > Capacity is optimal. > Airport charge includes not only the social marginal cost but price markup as well. > Ticket price paid by passenger is higher than social marginal cost > Capacity is too large. > Congestion toll and price markup in airport charge are higher. > Ticket price paid by passenger is even higher Social Welfare Maximization w/ Break-Even Constraint > Capacity and airport charge are weighted averages of the other two cases with shadow value of budget constraint being the weight > Implying capacity is optimal > Capacity and airport charge are weighted averages of the other two cases with shadow value of budget constraint being the weight

Oligopoly

Monopoly

> Capacity is too large. > Congestion toll and price markup in airport charge are even higher. > Ticket price paid by passenger is the highest

> Capacity and airport charge are weighted averages of the other two cases with shadow value of budget constraint being the weight

Conclusions and Discussions

Brueckner (AER 2002) assumed continuously differentiated passengers in travel preference enabling explicit separation of peak and off-peak periods Instead, the current paper considers capacity explicitly in a model which contains monopoly, oligopoly, and competitive market in a universal framework Despite such stark difference in the model setting, current paper shares the most relevant result with Brueckner that monopolist fully internalizes the congestion externality together with a negative term related to its market power

Conclusions and Discussions

Budget constrained airport is a weighted-average case of the welfare-maximizing and profit-maximizing airports with the shadow value of the constraint being the weight. Such relationship described in this paper between these three types of airports is analogous to that obtained in Zhang and Zhang (JUE 2003) which analyzed the effects of concessions on airport charges and capacity for the welfare-maximizing, profit-maximizing, and budgetconstrained airports.

Conclusions and Discussions

Comparative static results regarding carriers choice problem at stage 2 can be alternatively derived with an explicit use of the information at the optimum.

Pricing, Capacity, and Construction Boundary of a Congestible Highway with an Elastic Demand: Social Optimum, Second Best, Privatization, and Vertical Disintegration

Yuichiro Yoshida Associate Professor National Graduate Institute for Policy Studies

Research Objective To nd socially optimal capacity and pricing of a congestible Highway with and without vertical disintegration Using the one-dimensional city model with land use being ignored; thus partial equilibrium There exist alternative transportation modes; the demand is elastic Analyzes the impact of privatization when a transport company has market power Also analyzes the impact of disintegration of ownership and operation

Background Optimal highway policy have been analyzed in conjunction with the optimal land use Mills and de Ferranti (1971) showed optimal allocation of land for transportation Hochman (1975) showed that competitive equilibrium with a toll equivalent to congestion externality decentralizes the optimum in M&F However, these had a set of restrictive assumptions and functional speci cations xed lot size and xed number of commuters congestion is zero-degree homogeneous to ow and capacity

Background Oron, Pines, and Sheshinski (1973) showed the socially optimal congestion pricing by incorporating congestion to the Mills and Muth s model, however, road width was xed Akai, Fukushima, and Hatta (1997) elaborated on OPS model by endogenizing the land use for transportation and showed that perfect competition results in socially optimal allocation of land without the existence of transit authority The AFH model is a general equilibrium model with perfectly competitive transportation sector Current paper extends these existing literature by assuming an imperfectly elastic demand for transportation, though it is a partial equilibrium model as it disregards the land use for housing

Introduction The paper starts out by analyzing vertically integrated ownership and operation First it solves the social optimum, and shows that the standard optimality conditions implied by transport literature still holds even if the model is explicitly spatial Next it analyzes the impact of privatization and shows that the toll is higher than the marginal congestion cost as much as monopolistic markup

Introduction Then the paper analyzes the cases of vertical disintegration First case is when two separate private companies owns and operates the highway Second is the case of public ownership under private operation It is shown that private-private case results in double marginalization In the public-private case the public owner undoes all the distortions caused by the private operator These are strongly analogous to the results obtained by Zhang and Zhang (2006) for the case of Airport

Basic Model Highway stretches out from CBD in one direction Commuters reside in this one-dimensional city along this highway and must commute to CBD There are imperfectly substitutive transportation modes other than this highway; demand for this highway transportation services is elastic to price We analyze only the inbound trip on this highway

Basic Model c ( f (x) , w (x)): Cost of transporting one individual a unit distance at a location x on a highway f (x) : total ow at x w (x) : width of highway at x Assume that c/f > 0 and c/w > 0, and that 2 c/f 2 > 0, 2 c/w2 > 0 , and 2 c/f w < 0 Let x = 0 be the location of CBD and x = x be the city boundary

Basic Model The total cost for this highway is consisted of aggregate user cost and the construction cost:
x

[f (x) c ( f (x) , w (x)) + k (w (x))] dx

0

k (w (x)): the (instantaneous) highway construction cost at x (assume k has a xed part and k 0.) n ( p ( x)): the demand function for the highway service at x p ( x) : the user cost plus the toll, i.e.,
x

p ( x) =
0

[c ( f () , w ()) + t ( x dx )] x x

t ( x) is an instantaneous toll at x. Then note that we have f (x) = n ( p ( x)) .

Vertically Integrated Ownership and Operation: Social Optimum social optimum maximizes the net bene t
n( p( x))

B (n (p ( x)))
0

n () dn.

The maximization problem becomes as follows:

x w( x) ,t( x)

max

[B (n ( p ( x)))
0

f (x) c ( f (x) , w (x))

k (w (x))] dx

s.t.

f (x) = n ( p ( x)) p (x) = c ( f (x) , w (x)) + t ( x)

together with boundary conditions that p (0) = 0 and f (x) = 0. The corresponding Hamiltonian is H = B (n ( p ( x))) f (x) c (f (x) , w (x)) k (w (x)) + 1 ( n ( p ( x))) + 2 (c ( f (x) , w (x)) + t ( x))

10

Vertically Integrated Ownership and Operation: Social Optimum FOCs are Hf = c ( f, w) + f c c + 2 = f f 1 n (p) = 2 c =0 k (w) + 2 w
1

Hp = pn (p) c Hw = f w Ht = 2 = 0 for all x [0, x].

11

Vertically Integrated Ownership and Operation: Social Optimum Rearranging the above yields f c w = k (w) , c , f x [0, x] x [0, x ] .

t = f

Socially optimal capacity is such that the marginal cost and bene t of capacity expansion are equated at any point along the highway. Optimal (instantaneous) toll must be equal to the congestion externality cost at each location. If capacity of the highway is already predetermined, i.e., if the highway is already constructed, whether or not optimally, the optimal (second-best) toll is just as much as the congestion externality occurring at each location. This means that, in such a location where there is no congestion (even under no toll), the optimal toll should be zero.

12

Vertically Integrated Ownership and Operation: Social Optimum At the end of the highway x, we have k (w) = B (n ( p (x))) pn ( p) .

The end of the highway is such that the marginal cost of extending the highway is equal to the (instantaneous) consumer surplus that will be generated by such marginal extension of the highway. As the highway extends far from the CBD, the user cost p increases, and thus the consumer surplus becomes smaller. Once the consumer surplus becomes smaller than the xed cost of constructing a highway, it is the end of the highway, namely, x.

13

Vertically Integrated Ownership and Operation: Social Optimum Capacity gradient w is decreasing in x at all locations: n w = c ( f, w ) 2 c ( f, w) +f w f w 2 c ( f, w) f w2 k <0 x [0, x ] .

Given the slope of optimal toll capacity gradient is ambiguous in sign: t = n 2 c ( f, w) c ( f, w ) +f f 2 f 2 c ( f, w) w +f wf

There are two eects in t , negative and positive. The rst term captures the decrease in congestion due to smaller ow as x increases. The second is the increased congestion due to less capacity for a location farther from CBD.

14

Financing Viability of the Social Optimum Assume that c ( f, w) is 0 homogeneous in f and w, therefore c c f+ w=0 f w Using this in the rst-order conditions gives tf = k (w) w. Thus, for a smaller value of w, k (w) is greater than tf ; however, for a suciently large w, k (w) will be smaller than tf .

15

Financing Viability of the Social Optimum Given that the optimal capacity decreases from the CBD to the city boundary x, the toll revenue will exceed the construction cost in the neighborhood of CBD, and the subsidizing is necessary toward the end of the highway, under an optimal highway policy. The next step is to investigate the second-best problem where the transit authority must be nancially independent and thus faces a break-even budget constraint. The problem will be analyzed in two dierent cases where the capacity is determined endogenously and exogenously.

16

Second Best with a uniform-toll constraint Highway is still vertically integrated under the public ownership and operation TA maximizes the social welfare under a uniform-fare constraint: t ( x) = t x We rst assume that the transport authority does not face a break-even constraint Then we consider a case with the break-even constraint.

17

Second Best with a uniform-toll constraint TA s welfare maximization problem becomes:

x w( x) ,t

max

[b ( n ( p ( x)))
0

f (x) c ( f (x) , w (x)) x

k (w (x))] dx

s.t.

f (x) = n ( p ( x)) x p (x) = c ( f (x) , w (x)) + t

where the boundary conditions are p (0) = 0 and f (x) = 0 as before.

18

Second Best with a uniform-toll constraint Corresponding rst-order conditions are c ( f (x) , w (x)) + f (x) c = f p ( x) n (p ( x)) 1 (x) n (p ( x)) = c c f (x) = 0 k (w (x)) + 2 (x) w w +
2

c f

(x)

1 2

(x) (x)

x x

2 0

(x) dx = 0.

19

Second Best with a uniform-toll constraint Transversality conditions are [b ( n ( p ( x))) f (x) c ( f (x) , w (x)) k (w (x))] 1 (x) n ( p ( x)) + 2 (x) [c ( f (x) , w (x)) + t] = 0 1 (0) = 0 2 (x) = 0 with x being free Using f (x) = 0 and
2

(x) = 0, the rst equation simpli es to

1

b ( n ( p ( x)))

(x) n ( p ( x)) = k (w (x)) .

20

Second Best with a uniform-toll constraint With a break-even constraint, TA s problem becomes:

x w( x) ,t

max

[b ( n ( p ( x)))
0

f (x) c ( f (x) , w (x)) x

k (w (x))] dx

s.t.

f (x) = n ( p ( x)) x p (x) = c ( f (x) , w (x)) + t

x x

t
0

f (x) dx =
0

k (w (x)) dx

where the boundary conditions are p (0) = 0 and f (x) = 0 with x being free as before, for which the Lagrangean is
x

L =
0

b ( n ( p ( x)))
x

f (x) c ( f (x) , w (x))

x

k (w (x)) dt

+
0

k (w (x)) dx

t
0

f (x) dx

where

is the Lagrangean multiplier. 21

Second Best with a uniform-toll constraint The corresponding rst-order conditions are c ( f (x) , w (x)) + f (x) c = f p ( x) n (p (x)) 1 (x) n (p ( x)) = c c f (x) = k (w (x)) + 2 (x) w w +
2

c f

(x)

1 2

(x) + t (x) x

x x

k (w (x))
x

2 0

(x) dx =
0

f (x) dx.

22

Second Best with a uniform-toll constraint Transversality conditions are the same as before: [b ( n ( p ( x))) f (x) c ( f (x) , w (x)) k (w (x))] 1 (x) n ( p ( x)) + 2 (x) [c ( f (x) , w (x)) + t] = 0 1 (0) = 0 2 (x) = 0 Again, using f (x) = 0 and b ( n ( p ( x)))
2

(x) = 0, we have
1

(x) n ( p ( x)) = k (w (x)) .

23

Simulation of the Second Best: User Cost Function User Cost c is as follows: c ( f (x) , w (x)) = 1+ f (x) w (x)

c is homogeneous of degree zero in f and w measured in terms of the number of vehicles. is the value of in-vehicle time is the time it takes to travel a unit distance in a free ow = 62.86 yen per minute = 6 at free ow speed of 100km/hour for each 10km segment = 0.48 and = 2.82

24

Simulation of the Second Best: Capacity Cost Capacity cost k (w) is the opportunity cost of highway infrastructure and maintenance for each 10km segment for one day: k (w) = k0 + k1 w Parameters are k0 = 486400 k1 = 0.0000104 = 2.5. These values are determined as follows Historical data yields the highway construction cost as 3.72 billion yen/km.4% opportunity cost of capital gives 148.8 million yen/km-year of construction cost Maintenance cost is 28.75 million yen/year, sums up to 486,4000 yen/km or 4.864 million yen per segment. 25

Average capacity of highways in Japan in 2006 is 446,400 vehicles/day. Fixed cost k0 is 10% of entire cost k (w) at this point, as well as an assumption that = 2.5, we get k1 .

26

Simulation of the Second Best: Demand Demand function at location x is linear such that its inverse is: p ( x) = p0 p1 n ( x) , p0 , p1 > 0.

p0 and p1 dier at each location in a way the actual demand n and price p yields the price elasticity of 0.6e get k1 .

27

Simulation Case 0 : The Benchmark Case Capacity and pricing are set at the actual level Capacity w is 72, 000 vehicles/day upto the 9th segment and is 48, 000 vehicles/day for segments from 10th to the 20th Toll is 24.6 yen/km or 246yen per one segment (Case 0-1) Social welfare is computed as 235.0 million yen per day Loss is 22 million yen per day Marginal external cost (MEC) in the rst segment at CBD is as high as 1, 011 yen, while that at the end of the highway is just 4 yen Alternatively, toll is set for the rst 4 segments as 29.5 yen/km while for 5th and beyond it is 24.6 yen/km (Case 0-2) Social welfare is computed as 236.6 million yen per day with the loss of 9 million yen per day

28

Simulation Case 1: The First Best No break-even constraint Social welfare is as much as 309 million yen per day Optimal capacity in the rst segment is 79, 507 vehicles/day, and for the last segment 15, 480 Toll is (=MEC) for the rst segment is 562 yen, and 92 yen for the last segment Pro t is 130 million yen per day.

29

Simulation Case 2: The Second Best With the uniform-toll constraint No break-even constraint. Second-best toll is computed as 378 yen Social welfare is 307.6 million yen Capacity is slightly larger compared to the rst best for all segments, yet, it is much smaller than the actual capacity Pro t is 144 million per day

30

Simulation Case 3: The Third Best Both the uniform-toll constraint and break-even constraint are imposed Social welfare is 297.2 million yen Third-best toll is lower than in the second best, and is 172 yen. With this low price, ow is larger and to accommodate this large capacity is larger near the CBD. ow,

MEC is 661 yen for the rst segment while it is 87 yen at the end of the highway

31

Simulation Case 4: Uniform Toll under Actual Capacity without Break-Even Constraint Capacity is at the actual level: 72,000 vehicles/day for the rst 9 segments and 48,000 vehicles/day for 10th through 20th segments Optimize the level of uniform toll given the capacity without a break-even constraint The optimum toll is 321 yen per segment, which is about 30% higher than the actual toll of 246 yen/segment Social welfare is greatly reduced to 236.5 million yen per day, only 1.5 million larger than that in the benchmark case (case 0-1) There is a small surplus of 23 million yen/day

32

Simulation Case 5: Uniform Toll under Actual Capacity with Break-Even Constraint Both the uniform-toll constraint and break-even constraint are imposed Social welfare is 236.1 million Toll level is 281 yen/segment, which is close to the actual level, re ecting the fact that this setting of uniform toll with break-even constraint under the actual capacity is the closest scenario to the actual situation given capacity Case 0-2 has higher welfare compared to cases 4 and 5, due to its dierentiated toll schedule

33

Simulation Case 6: Location-Varying Toll under Actual Capacity without Break-Even Constraint Lastly, we analyze the situation where the highway authority optimizes the toll level at each segment given the actual capacity, without facing the break-even constraint The toll is 754 yen for the rst segment, while it is only 4 yen at the end Travel demand measured as the trac ow at the CBD is lower: though trac volume near the end of highway is increased, suppressed travel demand near the CBD outweighs such increase Social welfare in this case is only 241 million yen/day. Despite the optimization of toll at each location, the welfare gain was almost negligible. Loss is just 4 million yen per day.

34

Privatization: Pro t Maximization Highway corporation s pro t maximization problem is

x w( x) ,t( x)

max s.t.

[f (x) t ( x)
0

k (w (x))] dx

f (x) = n ( p ( x)) p (x) = c ( f (x) , w (x)) + t ( x) ,

together with boundary conditions that p (0) = 0 and f (x) = 0. Solving the above problem yields the following: f (x)
x 0

c = k (w (x)) w
x 0

x [0, x ] x [0, x ]

) t ( x dx =

p ( x) c dx + f () x En:p f

where En:p is the elasticity of demand n with respect to price p.

35

Privatization: Pro t Maximization Toll charged by the pro t maximizing rm is higher than the social optimum as much as the monopolistic markup. Yet, the pro t maximizing condition for the highway capacity has the same expression as that of social optimum. Hence, ow f is smaller at any location than the social optimum and the capacity is smaller. As the competition increases, En:p becomes large. When the transportation market is perfectly competitive, En:p becomes in nity and the pro tmaximizing toll is socially optimal. This result is similar to those pointed out by Knight and Akai, Fukushima, and Hatta

36

Privatization: Pro t Maximization The transversality condition at the end of the highway is k (w (x)) = p ( x) n ( p ( x)) . En:p

Therefore, cost of extending the highway for a unit distance is equal to the sum of the monopolistic mark-up paid by the users at x The above can be alternatively written as
x

k (w (x)) = p ( x)
0

c ( x) + f (x)

c ( f (x) , w (x)) dx n ( p ( x)) . f (x)

In other words, a pro t-maximizing rm will construct a highway up to a point at which the price paid by the passenger at such point minus the cost and externality of transporting them to CBD is equal to the construction cost.

37

Privatization: Pro t Maximization If we can assume that the price elasticity of demand is constant, i.e., En:p is equal to some constant e say, for all p, then, t ( x) = f (x) c ( f (x) , w (x)) f (x) p (x) e x [0, x] ,

Toll charged by the pro t-maximizing rm is not zero even at the end of the highway x, unlike in the social optimum. Let m e/ ( e + 1) be the monopolistic mark-up multiplier such that m 1 is the rate of monopolistic mark-up, and m 1 = 0 (or, m = 1) in case of perfect competition (as e = ). Then the above equation becomes p (x) = m c ( f (x) , w (x)) + f (x) c ( f (x) , w (x)) f (x) x [0, x ]

implying that the rm will fully internalize the congestion externality, then multiply the full social marginal cost by the mark-up constant m. 38

Privatization: Pro t Maximization Again, for a similar reason as in the social optimum, the sign of t (x) is ambiguous while w (x) is negative for all x: t =m where n w = n 2 1 m c 2 c +f 2 f f 2 c + f + 1 wf 1 m c w w

c 2 c +f w fw 2 c f 2 w

k <0 x [0, x] .

t converges to that of social optimum as m approaches one.

39

Vertical Disintegration: Separated Private Ownership and Operation First the owner decides on the capacity of the highway and sets the fee of leasing the highway infrastructure to the operating rm Highway leasing fee is assumed to be proportional to ow volume at each point x, i.e.,
x

f (x) u ( x) dx
0

where u ( x) is the unit leasing fee at location x per-passenger Given the capacity and leasing fees, the operator sets the toll at each point along the highway We solve this case backward from the operator s pro t maximization problem in the second stage

40

Vertical Disintegration: Separated Private Ownership and Operation Operator s pro t maximization problem becomes
x

max
t( x) 0

[f (x) t ( x)

f (x) u ( x)] dx

s.t.

f (x) = n ( p ( x)) p (x) = c ( f (x) , w (x)) + t ( x)

together with boundary conditions that p (0) = 0 and f (x) = 0, and x given in owner-decision stage Solving above shows that the operator internalizes the congestion externality and charges monopolistic mark-up, on top of the leasing fee u (x) .:
x x

) t ( x dx =
0 0

p ( x) c ( f () , w ()) x x + + dx f () x En:p f () x

) u ( x dx
0

x [0, x ]

41

Vertical Disintegration: Separated Private Ownership and Operation Assuming again that En:p = e for all (n, p) for analytical tractability we have p (x) = m c ( f (x) , w (x)) + u ( x) + f (x) c ( f (x) , w (x)) f (x) x [0, x]

Operator internalizes the marginal congestion cost and charges full marginal cost including the leasing fee, then further multiply it by m, the monopoly mark-up multiplier.

42

Vertical Disintegration: Separated Private Ownership and Operation Given this behavior of the operator, the maximization problem for the owner is set up as
x u( x) ,w( x)

max

f (x) u ( x)
0

k (w (x)) dx c ( f (x) , w (x)) + u ( x) f (x)

s.t.

f (x) =

n ( p ( x))

p (x) = m c ( f (x) , w (x)) + f (x)

together with boundary conditions p (0) = 0, f (x) = 0, and x being free.

43

Vertical Disintegration: Separated Private Ownership and Operation Solving the owner s problem gives the optimal leasing fee gradient c ( f, w) + (m u = 2f f and a capacity condition k (w) = 2 c ( f, w) c ( f, w) +f f w wf
2 c ( f, w) 2 c ( f, w) 1) c ( f, w) + f + mf f f 2

By substituting the optimal lease fee back into the optimal toll function we get p =m
2 2 c ( f, w) 2 c ( f, w) c ( f, w ) + 3f +f f f 2

for all x [0, x], and the transversality condition x c 1 k (w (x)) = p ( x) c+f + u dx n ( p ( x)) . m f
0

x [0, x] .

44

Vertical Disintegration: Separated Private Ownership and Operation Pro t-maximizing lease fee equation reveals the double marginalization by the owner. The owner charges twice the marginal congestion cost while the operator charges once. The second term captures the monopolistic mark-up by the owner, which is charged on top of these congestion charges. These items in the leasing fee is exogenous to the operator, and thus passed on to the users of the highway entirely, and even exaggerated by the operator as much as the mark-up. Leasing fee charged by the private owner is always positive, however, as we will see in the following section, the optimal leasing fee is non-positive under the private operator. This implies that the separated private ownership and operation is clearly suboptimal. Price paid by the users not only includes three time the congestion externality, but also multiplied by monopolistic mark-up twice, once by the owner and again by the operator. 45

Vertical Disintegration: Separated Private Ownership and Operation With k being non-negative and 2 c/wf being negative, the capacity condition implies that for a given ow f , the highway capacity is bigger than that in vertically-integrated cases or public-ownership case. Length of the highway is de ned implicitly in the transversality condition Terms in the square bracket on the right-hand side of the transversality condition above is the monopolistic mark-up at x. At the end of highway, the cost of extending the highway for another unit distance is equated to the sum of monopolistic mark-up paid by those users at x, divided by m. Such revenue is divided by m, as the owner gets only 1/m of what the operator actually charges to the users. As m approaches 1, k (w (x)) reaches zero, indicating that highway will not be constructed unless there is some market power under the private ownership.

46

Vertical Disintegration: Public Ownership under Private Operation Consider a case where the operation of highway is still private, but the owner is public who maximizes social welfare. Here again we de ne the social welfare as total bene t minus total cost. The problem for this public owner of the highway becomes
x w( x) ,u( x)

max

[b (n ( p ( x)))
0

f (x) c ( f (x) , w (x))

k (w (x))] dx

s.t.

f (x) =

n ( p ( x)) c ( f (x) , w (x)) + u ( x) f (x)

p (x) = m c ( f (x) , w (x)) + f (x) where b ( n ( p ( x)))

0

n( p( x))

() dn n

together with boundary conditions p (0) = 0, f (x) = 0, and x being free.

47

Vertical Disintegration: Public Ownership under Private Operation Solving the problem we obtain u ( x) = k (w (x)) = 1 m f (x) 1 c ( f (x) , w (x)) + f (x) c ( f (x) , w (x)) f (x)

c ( f (x) , w (x)) w (x)

and the transversality condition k (w) = b ( n ( p)) pn (p) .

Substituting u ( x) in the state equation for price gives p (x) = c ( f (x) , w (x)) + f (x) c ( f (x) , w (x)) . f (x)

48

Vertical Disintegration: Public Ownership under Private Operation Public owner will set the leasing fee u ( x) in such a way that resulting toll that the private operator sets equals the congestion externality at each location Such level of u ( x) is negative whenever there is a mark-up. Regardless of the size of market power, the public owner will construct the highway up to the point where the cost of extending a unit distance is equal to the additional net bene t received by those who at the margin. The transversality condition is the same as that of the social optimum in the vertically-integrated case, thus the length of the highway is socially optimal Public ownership can fully replicate the social optimum even under the private operation. De cit of the owner may well be nanced through the lump-sum payment by the operating rm, and it is important that the owner monitors the ow volume at each location on the highway. 49