OPTIMAL FARES AND PRODUCT QUALITY: THE CONTRIBUTIONS OF DOUGLAS AND MILLER While the discussion so far has acknowledged that CAB regulation of interstate plane fares has some effect on product quality (flight frequency and load factor), it has taken no explicit, quantitative account of these effects. The first work to analyze these dimensions of air transportation quantitatively was that of Douglas and Miller, published in 1974.75 This work attempts, in a most ingenious way, to quantify the benefits and costs of the quality changes which CAB regulation has brought about. The basic idea behind their methodology is that lower load factors allow more convenient service (in ways described below), but cost the traveler more in fares. Their model trades off the value of this convenience to the traveler against the cost of providing it. The Douglas-Miller methodology They assume initially that every traveler has a "most preferred" time of departure. To the extent that he is unable to leave exactly when he would most prefer, he is subject to a "schedule delay" equal to the difference between the time he would prefer to depart and the time he does depart. This schedule delay is made up of two components. First, it is unlikely that any one flight will be scheduled exactly at the desired departure time, even if the traveler can get reservations on that flight. The difference between desired departure time and the departure time of the nearest flight is termed frequency delay. Furthermore, the flight closest to the desired departure time may be booked. The average time incurred waiting because the nearest flight is book is termed stochastic delay. Schedule delay is simply the sum of frequency delay and stochastic delay. Douglas and Miller not only define the concept of schedule delay, but they also provide estimates of how much schedule delay is likely to occur on a given route per passenger (averaged over the year) as a function of the amount of traffic on the route, the size of aircraft used, and the number of flights (assuming a given seating configuration). Holding the number of passengers on a route and the capacity of an aircraft constant, their method provides an estimate of expected schedule delay per passenger as a function of load factor. Now, a lower load factor costs more to provide in terms of operating costs, but it provides additional benefits in terms of reduced schedule delay time. The point of Douglas and Miller's analysis is to estimate and weigh these benefits and costs quantitatively. To do so, it is necessary to evaluate in money terms both the costs of varying load factors and the benefits of varying schedule delay. The costs of offering various load factors and flight frequencies for a given trip, plan type, and seating configuration can most certainly be estimated, and Douglas and Miller do so. However, it is much more difficult to evaluate schedule delay time, that is, to determine the amount which the average passenger would be willing to pay to leave a minute (or hour, or whatever) closer to his preferred departure time. Douglas and Miller assume that the typical passenger would value his schedule delay time at somewhere between $5 and $10 per hour, i.e., 75 "Economic Regulation of Domestic Air Transport," chapters 5 and 6. 76 they assume these as alternative values for their calculations. These assumptions are probably somewhat on the high side, inasmuch as they are based on econometric estimates of the value of in-vehicle travel time, wherein the traveler is limited in the use he can make of that time. Schedule delay time, on the other hand, can be used more or less pleasurably and/or productively because it does not have to be spent In a vehicle or even waiting at the airport." For this reason, their results based on a $5 per hour value of schedule delay time are more likely to be valid than their estimates based on a value of $10.78 The output of the Douglas and Miller model consists of both optimal load factors and optimal fares, as a function of route, plane type, and seating configuration." The optimal fares are based on costs estimated in turn from a cross section of trunk airlines. Results of the Douglas-Miller model-Optimal fare level On the basis of their calculations, Douglas and Miller provide an estimate of the welfare loss from CAB regulation, as of 1969. Their results are worth quoting in full: During 1969, . . . domestic trunk-line revenue totalled $6,514 million. Had the price-quality options been at optimal levels consistent with a $10 per hour evaluation of passenger delay time, total passenger revenue would have been approximately $6,148 million; for an optimal configuration reflecting a $5 per hourpassenger evaluation of delay time, revenue would have been approximately $5,976 million. In other words, during 1969 passengers paid excess fares, ranging from approximately $366 million to $538 million. However, for this additional price, passengers purchased reductions in delay time-valued at $118 million in the $5 per hour configuration consistent with a $10 per hour waiting time evaluation, and approximately $182 million in the $5. per hour configuration This leaves a 1969 deadweight welfare loss which (given the assumptions) ranges between $248 million and $356 million.80 After quoting this, it is necessary to point out also that Douglas and Miller qualify their welfare loss results by asserting that they are "just not that important" relative to the basic notion they introduce that service quality considerations are important to calculations of the welfare impact of airline regulation. The point remains that service quality can be "too high"-that it reduces overall schedule delay below levels for which customers, given the choice, would be willing to pay. Furthermore, Douglas and Miller's numerical results are of considerable interest, because they imply a welfare loss from CAB regula 76 The demand studies from which Douglas and Miller get their time values are ArthurDe Vany, "The Revealed Value of Time in Air Travel," (Review of Economics and Statistics 56 (February 1974), pp. 77-82; and S. Brown and W. Watkins, "Measuring Elasticities of Air Travel from New Cross-Sectional Data," delivered at annual meetings of American Statistical Association, 1971. 77 If a passenger makes reservations, he knows when his plane will leave, and need only go to the airport just before it does. Thus, in the case of airline schedule delay. "waiting" simply amounts to changing one's plans so as to do something other than traveling at the most desired departure time. 78 Furthermore, if plane fares are currently higher than they optimally would be, it is more or less certain that if fares were lowered, new travel by those with lower time values would be induced to travel, drawing the average value of schedule delay time down further. As a result, it will achieve inappropriate results to use time values estimated only for those currently traveling by plane. 79 As route density rises, optimal load factor rises also, because as route density goes up, the extra benefits of frequency do not rise proportionally to the number of passengers. A longer haul tends to reduce the optimal frequency. For the reasons behind this, see the subsection on optimal fare structure below. 80 "Economic Regulation of Domestic Air Transport," p. 72. The welfare loss which they calculate is lower with a higher value of time because the low load factors which the CAB has brought about are most convenient for travelers with high time valuesthose willing to pay a high fare with the corresponding reward of easier reservations and less schedule delay. 81 tion of only 4 to 6 percent of total fares charged, whereas previous studies of the welfare costs of CAB regulation for the late 1960's and early 1970's would put the number at anywhere from 20 to 55 percent of interstate fares charged.82 To some degree, this difference stems from the fact that Douglas and Miller have subtracted the benefits of improved service quality made possible from CAB regulation from the costs in excess fares. However, that does not explain the whole difference, for their results imply that optimal fares are no more than 6 to 9 percent lower than actual ones on CAB-regulated routes. The main difference between Douglas and Miller's welfare loss estimates and those implied by previous studies would seem to stem from differences in measurement of costs and fares. Let us consider measurement of fares first. To measure actual fares, Douglas and Miller use not official fares, but instead revenue yields, taking account of discounts and the like to the extent that they are used. This has a problem, however, for the restrictions attached to discount fares make the trips attached to them inferior products, as discussed above. Therefore, it is at least as reasonable to compare actual regular fares on a given route with optimal ones, assuming that the "optimal" fares would have no strings attached (and indeed, the much lower regular fares available in intrastate markets do have no strings attached, except for the rock-bottom night and weekend fares in Texas). More importantly, it appears that Douglas and Miller have overestimated airline costs, at least relative to what they are on the California and Texas routes. For example, their cost estimates project that in 1971, the year for which their costs are calibrated, PSA would lose over $2.50 per passenger at a 100 percent load factor, and over $6 per passenger, or 40 percent of the fare charged, at a more realistic 70 percent load factor.83 Details of these comparisons, plus a more detailed critique of Douglas and Miller's cost estimates, are contained in section II of the appendix to this study. Douglas and Miller theory: Optimal fare structure Douglas and Miller's model analyzes not only the optimal fare level, but also the optimal structure of fares, i.e., the way they should vary by length and level of traffic for a given route. This requires further explanation. As previously stated, their model trades off the benefits of additional flight frequency on a given route (with a given flow of passengers per year) against the costs of that additional flight frequency.84 But the benefits and costs of flight frequency are not independent of trip length and route density (i.e., flow of passengers per unit of time 81 That is, total revenues in 1969 were $6.148 million, and Douglas and Miller's estimates of the costs of regulation were $243-356 million, or 4-5 percent of the total. 82 For example, my own estimates of the welfare costs of CAB regulation on the sample studies of 30 routes in 1968 was 20-90 percent of the efficient fares. This is equivalent to saying that the welfare loss is 17-47 percent of existing fares (1/1.2-.83, and 1-.83= .17: similarly, 1/1.9-.53, and 1-.53=.47). These calculations are based on Tables 2-7, p. 24 of Douglas and Miller. Further details of these calculations are provided in Section II of the appendix to the chapter of the present work. 84 This analysis is done holding number of passengers on a route constant. In general, a higher fare does not necessarily result in a higher frequency, despite the fallacious assertions of many that that is the case. The effect of a change in fare on flight frequency depends on the change on the number of passengers as a result of the fare change. traveling on the route). More specifically, as the length of an airline flight increases, the costs of providing additional flights increase as well. The benefits of this additional frequency, however, do not increase though in-vehicle travel time increases with length of flight, schedule delay time is a function of frequency, load factor, and plane size alone. Therefore, as the length of haul (and hence the cost of a flight) increases, it will generally be most efficient to trade off some of the higher cost of flights for more schedule delay time. In other words, as flights get longer in haul, it will pay to reduce flight frequency because long-haul flights are costlier than short-haul ones, on a per-trip basis. Therefore, optimal fares will entail higher load factors on longer hauls. Before the DPFI, CAB fares did just the opposite: the breakeven load factor declined as length of haul rose. 85 Another element of Douglas and Miller's theory of optimal fare structure relates to route density. As route density (again, the number of passengers traveling on a route per unit of time) increases, one would expect flight frequency to rise, as well. But Douglas and Miller show that it is not efficient to increase flight frequency proportionally with the number of passengers traveling on a route. This is because of the indivisibility of aircraft, i.e., on low density routes, aircraft cannot be made smaller than a certain size (say, 110 coach seats) without a sharp increase in operating costs. Maintaining an optimal frequency on a low-density route may thus require operating 110 seat planes at low load factors. As route density rises, the optimal strategy is to fill up more seats, so that optimal flight frequency does not rise proportionately with route density. Given that the benefits of flight frequency do not increase proportionally with passengers, the optimal load factor should be expected to rise with route density. Currently, all CAB fares are based on a 55 percent load factor, standard, independent of route density and of length of haul.86 Because of this, CAB regulation imposes inefficiencies from fare structure in addition to those stemming from fare level. Douglas and Miller's welfare loss estimate includes the costs of nonoptimal fare structure with respect to length of trip, but not with respect to route density. In subsequent work, however, based on Douglas and Miller's methodology, M. Pustay has attempted to estimate the cost of nonoptimal fare structure, as well as nonoptimal fare level.87 His estimates will be considered later in this section. Overall, Douglas and Miller's methodology of estimating optimal fares and service qualities is a most important contribution. But we shall attempt here to get estimates of the welfare loss from CAB regulation incorporating their model of schedule delay, but based on a cost model calibrated to fit more accurately the costs and fares achieved in the less-regulated environments of California and Texas, and to get at least a rough idea as to what additional costs are imposed by non 8 This is equivalent to saying that the ratio of fare to cost at a constant load factor rose with distance. That this was so may be seen from Keeler, "Airline Regulation and Market Performance." Table 8. Alfred E. Kahn, "The Regulation of Air Transportation-Getting From Here to There" in Regulating Business The Search for an Optimum, Institute for Contemporary Studies. p. 48 (1977). 87 "The Effects of Regulation on Resource Allocation in the Domestic Trunk Airline Industry," chapter 3. optimal structure in CAB fares not accounted for in Douglas and Miller's published work. Before doing that, however, it is important to consider another issue. The argument for deregulation hinges implicity on the assumption that competition will generate an optimal combination of fare and service quality. It is now time to examine the evidence as to whether this is so. THE EXISTENCE AND OPTIMALITY OF COMPETITIVE MARKETS IN As previously stated, competitive markets can fail to achieve economically efficient results because of three broad types of problems: monopoly power at the firm level caused by increasing returns to scale, externalities at the market level, and imperfect information about the market on the part of consumers and/or producers. It is worth considering the likelihood of each of these problems occurring in the trunk airline industry. Returns to Scale for Trunk Airlines Many statistical studies have been done by economists investigating the extent of increasing returns in the airline industry.88 The methodologies differ considerably from study to study, but the basic question asked is the same in each case: what happens to airline operating costs per unit of output as firm output rises (and as firm size adjusts these changes in output)? The most common way to answer this question is via statistical analysis of a cross section of firms, measuring how costs vary as one goes from small firms to large firms. In the simplest such studies, output is measured with only one variable, and it is possible to determine scale economies visually by way of a "scatter diagram," i.e.,. a diagram plotting firm costs per unit of output (average costs) against output. If average costs are constant with output, then there are constant returns to scale, and there is no reason to expect market failure, for reasons discussed above. Similarly, if average costs increase with firm output, that is similarly consistent with efficient, competitive markets. However, if average costs decline with firm output,. that is a sign of increasing returns to scale, and under those circumstances one should expect market failure, for reasons discussed in section III. 89 One recent study of returns to scale in the airline industry, by the CAB Special Task Force on Regulatory Reform, has measured returns to scale in a two-dimensional, graphic way, and the evidence it finds is worth showing here as well.90 88 Among the studies in footnote 1 which considered returns to scale in the airline industry are Caves (chapter 3), Keeler (indirect costs only, pp. 406-411), and Douglas and Miller, chapter 2. Also considering scale economies implicitly (by comparing efficiency measured in various ways for large firms and small firms) are Jordan, mentioned in footnote 1, and Robert J. Gordon, "Airline Costs and Managerial Efficiency," in J. R. Meyer, Editor, Transportation Economics: A Conference (New York, National Bureau of Economic Research, 1965), pp. 61-94. Finally, Pulsifer, et al., measure scale economies, as discussed below in the text. Sa It will be recalled that destructive competition and natural monopoly occur only when marginal cost is below average costs. If average cost is constant or increasing, then the cost of the last unit of output will always be equal to or greater than the average cost of all units, so marginal costs will be equal to or above average costs. 90 Pulsifer, et al., pp. 102-107. |