Ci;ostsand Scale of BusServices Public Disclosure Authorized

World Bank Staff Working Paper No. 325

April 1979 Public Disclosure Authorized Public Disclosure Authorized

The views and interpretationsin this document arethose of the authors and should not be attributed to the World Bank,to its affiliated organizations,or to any individual acting in their behalf.

Preparedby: A.A. Walters Urban ProjectsDepartment Copyright © 1979

Public Disclosure Authorized The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. WORLD BANK

Staff Working Paper No. 325

April 1979

COSTS AND SCALE OF SERVICES

This paper challenges the conventional wisdom that large in large organizations with subsidies to produce optimum frequencies are the best arrangement for urban road passenger . It is shown that in theory small buses are often appropriate, giving the best frequencies and speeds and suitably low average passenger waiting times. Direct observations of services in a number of cities tentatively demonstrate the veracity of the theory. Furthermore the best institutional organization is not the large firm or municipal authority, but the small firm, often the owner/driver. There is no case for any substantial subsidy for appropriately organized urban bus transport.

Prepared by: A. A. Walters Urban Projects Department

Copyright O 1979 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. ACKNOWLEDGEMENTS

In writing this paper the author received valuable and constructive comments from H. Mohring, R. Kirby, A. Goldstein, J. Cracknell, G. Jacobs, H. Levinson, H. Dunkerley, A. Churchill, and G. Roth. Additional data were provided by C. Rees (on Kuala Lumpur) and R. Podolske (on Bangkok). Mrs. Betty Easter tirelessly prepared the manuscript and removed many inconsistencies and errors. - ii -

INTRODUCTION

Bus services are the most important mode of motorized transport in citi-s in developing countries. And, in contrast with the standard large buses of the municipalized or nationalized undertaking in the developed countr-ies9ci ties ir developing countries have evolved a wide variety of bus systems and institutional structures. The jitney, the shared taxi, the minibus i(12seats or so), the microbus (less than 12 seats) and the midibus (15-25 seats)--all exist, indeed flourish, sometimes alongside the standard bus which one sees in the cities of Europe and America. Similarly, while the standard large bus is often, though not always, operated by the nationalized concern, the small buses and taxis are normally owned and operated by their main driver.

The natural policy questions that emerge are: What is the best institutional form for urban bus services? Should they be nationalized or at least stringently regulated as the only way to "integrate" or "coordinate" bus services---orcan these functions be performed better in another way? Should there be a bus subsidy and if so in what form? What size should the buses be? Should there be a mixture of sizes as well as institutional forms?

For an informed discussion on such policy matters, one requires first an analysis of the principles involved. The concept of economies of scale in the bus business has been confused with the economics of the size of bus. Whether larger firms enjoy lower or higher costs is quite different from the issue of larger buses being more or less efficient than smaller buses. Also to be considered as part of the input of bus operation is the time spent waiting or traveling by passengers; therefore any optimization policy resulting from these variations in time inputs should be taken into account. It is clear that a free market system would take into account much of the time-of- passenger effects--and primarily by using a "small bus" that would give rise to profitable frequencies and trip times. The external economies of scale of increased frequency of bus services would be largely incorporated in private decisions, and thus there is no reason to suppose that, above some very small minimum size of firm, there are any substantial economies of scale either external or internal. On the contrary there are good reasons for supposing that unit costs will be higher for the large firm and that it will have less incentive to take into account the externality of passengers' time.

If these theoretical conjectures have any substance one would expect to observe in cities where there is a wide degree of freedom to develop appropriate vehicle size and institutional forms: (a) markedly small buses on the average, (b) small firms, (c) lower and better service, that is to say better frequencies, than in cities which have the traditional large bus in a municipal organization. Similarly where there is competition between the small bus operator and the large municipal organization, one would expect the small man to be the winner. These predictions are borne out in a survey of a number of cities in developing countries (and in the odd case of Belfast) where there is some freedom for private operators to enter the industry. - iii -

The main policy implication to be drawn from this discussion is that there is a prima facie case for giving private operators some substantial freedom to enter the urban bus business. The second part of this paper concen- trates on an evaluation of just such a policy change. In Kuala Lumpur the government allowed a substantial but limited number of to enter the industry during 1974-75. It is shown that very large benefits flowed from this policy change--amounting to nearly one percent of the incomes of residents of Kuala Lampur--and even larger benefits are achievable with a more liberal policy.

Finally, it is worth emphasizing that throughout the cities surveyed here, the municipalized or nationalized bus companies always had higher costs, although they had significant advantages granted by their concession. This is partly because in bus transport there seem to be significant economies of being small. The implication which needs to be explored is that free entry into urban bus transport is the best policy option for cities in developing countries. This would provide a fertile ground for the deployment of indivi- dual incentives, the growth of entrepreneurship and savings and investment, and for greater use of labor-intensive "appropriate technology." - iv -

CONTENTS

INTRODUCTION ...... ii

I. ECONOMIES AND SUBSIDIES IN BUS BUSINESS ...... 1

A. Economies and Externalities ...... 1 B. Profitable Adjustments of Bus Frequencies.....*..... 3 C. Bus Size, Waiting and Traveling Time ...... 5 D. A Simple Model of Bus Size...... 7 E. Regulation and the Consequences ...... 15 F. Case Studies ...... *.... 18 G. A Summary of Empirical Evidence ...... 24 H. Conclusion on Policy ...... 26

II. THE BENEFITS OF MINIBUSES: THE CASE OF KUALA LUMPUR ..... 28

A. The Minibus Policy...... es ...... 28 B. Alternative Regulatory Systems and Effects*...... a.... 30 C. Estimates of Benefits...... 33 D. Conclusions on Policy...... **...... 38

Tables

1.1 A Summary of Status of Minibus 25

Figures

1.1 Manning Scales 11 1.2 Comparison of Trips: Buses/Minibuses 16 2.1 Demand for Passenget Trips: Buses/Minibuses 30 2.2 Demand and Supply of Passenger Trips by Minibus under Different Regimes of Regulation 31 2.3 Demand for Minibus Trips 34

Annexes

1 Costs and Sizes of Buses 2 Costs of Bus and Minibus 3 Profits of Minibuses I. ECONOMIES AND SUBSIDIES IN BUS BUSINESS

A. Economies and Externalities

1.01 It is sometimes taken as axiomatic that there are substantial economies of scale in the business of providing bus services. And such economies have been adduced as a rationalization for the argument that urban buses should be subsidized. The most sophisticated version of the argument has been advanced by Mohring (1972). He suggests that additional bus services--characterized by increasing the frequency of buses of a fixed size on a particular route--will inter alia reduce passenger waiting time. This reduction in average costs means that marginal costs (in the long run) are less than average costs, and consequently long-run marginal cost pricing would require a subsidy. 1/ The problem has been analyzed further by Nash (1978); he asks what would be the optimum mark up to cover costs.

1.02 One preliminary point is that economies of scale are usually relevant for the firm. Mohring's formal analysis, however, applies to the route and assumes implicitly that the firm or municipalized undertaking is an aggregate of such routes. The bus firm is then assumed to enjoy constant returns to scale with respect to the expenses that it incurs (excluding passenger time) in providing hours of bus service. 2/ The increasing returns arise solely from passenger time effects. Nash, on the other hand, examines economies of scale "with respect to vehicle size" (my italics) and concludes that they are obvious. 3/ Nash explicitly rejects the argument that one might operate smaller vehicles: "if traffic were less dense, the cost savings would probably not be large enough to affect the results significantly." Thus both Nash and Mohring assume away the traditional issues about size of firm and economies of scale and suppose that the only limit to the size of bus is due to technological or safety constraints; otherwise, the bigger the bus, the better. And both consider a monopoly supplier of bus services at least for the route and probably for the city.

1.03 It is worthwhile stepping back a little from this monopoly and reflecting on the consequences of an alternative unregulated system of free enterprise provision of urban transport. This seems so far from experience,

1/ Mohring's development includes also a congestion effect in the additional passengers slowing up the bus and existing passengers, and creating system effects by the reduction of bus miles. But that is a side issue in the general argument for a specific subsidy.

2/ The problem then tackled by Mohring is to discover the optimum frequency and stops and journey speed of these given bus hours--which will give rise to a number of passenger journeys/miles and hours of time both traveling and waiting.

3/ Footnote 3, page 84. "If it were not for (these economies) operators could offer each passenger a separate individually tailored service." -2- at least in the affluent West, that it will take some effort of the imagina- tion--but I think that some idealizedconcept is necessary to see what is foregone in the monopolized-big-bussystem.

1.04 Obviouslyby using big buses only, much variety is lost. There is enormous variation in the marginal valuations of time, comfort, reliability and many other characteristicsof . Evidence of this appears in the rather limited choice which regulationat present offers--privatecar, and bus ( boat and , etc.). If there are no costs of making contracts or deals, then the technology(the size and running speed of the vehicle, the route, the number and type of stops, the comfort level, etc.) would respond to the bids of passengers. In principle,anyone could reduce his waiting time to virtually zero by bidding high for a pickup--andthe existing passengersin the vehicle could decide whether they would be willing to delay their journey for an offer so high (assumingthat they had found it worthwhile to enter into a conditionalcontract with the operator).1/

1.05 In this idealizedworld, therefore,the costs to "other passengers" (or what Mohring calls the "own bus" costs) and the costs because the operator knows that stopping the vehicle will mean that he does less business (Mohring calls this the "system" effects) are both reflectedin the price system. There is no need for any special congestionlevy; it is already there.

1.06 Of course, each firm may be subject to rising costs after some critical level of output, due perhaps to the limited supply of entrepreneur- ship. Rising costs of the individualfirm, however, may be associatedwith a falling long-run average and marginal cost curve of the industry if there are external economies,that is, external to the firm but internal to the industry. This is the competitiveanalogue of Mohring's monopoly. To be eligible for a subsidy, the external economiesmust be technicaland not pecuniary. And in this case they are: the greater the quantity with a fixed size of bus, the lower the waiting time.

1.07 It is important to note that the Mohring assumption of a constant technology whatever the size exaggerates the external economies of scale effect. There are many ways of increasing frequency and many ways of varying waiting times in response to demand. The ideal offer procedure with passenger bid and operator-cum-passenger response outlined above is approximated in various practices which involve saving on the expense of information and contracting. Well-defined "express services" and "door-to-door" services would set up at the side of the mass transit slow stage services. Operators would have the incentive to adjust their technology to the demand and probably some classification of service would emerge. Vehicles would probably exhibit a standard schedule for stops and special stops and an hour-of-day tariff. And passengers who got into a specific kind of bus would know the probability of stopping and the approximate journey time compared with other services.

1/ Ignoring the "free rider" problem. Similar difficulties occur if there is a large number of passengers waiting at the stop; it is then a group decision. 1.08 In a free system, one may well conjecture that the average sized firm would be rather small if one were to judge by the few cases where there is free entry. Notwithstandingthe municipalizationor nationalization(or integration--whateverthat may be taken to mean) of bus services, the evidence is that internal economiesof scale are quickly exhausted. 1/

B. ProfitableAdjustments of Bus Frequencies

1.09 One interestingquestion is whether or not Coase's theorem would apply to the demand for the "quality"of the service which is reflected first in the reduced waiting time and second in the reduction in journey time. 2/ These are not externalitiesin the strict sense that they are not the subject of market transactions. In the ideal model they clearly are. Bus operators would not merely adjust their equipmentbut also their operating patterns to respond to profitable reductions in both waiting and journey times. In principle, they would produce reductions in time provided that the passengers' marginal valuation of the time savings exceeded the marginal costs of produc- tion. 3/

1.10 But this is the story for a particular firm acting individually. In particular, if the firm expanded its services,while other firms did not, there would be a miniscule increase in the frequencyof services and a small reduction in waiting time. 4/ If, however, all firms expanded their services together, there would be a significantreduction of waiting time (cet par) and so an increase in the marginal valuation of existing travel and probably some expansion of the kilometers traveled. Thus, it appears as a market failure-- each individualfirm finds it unprofitableto expand, but if everyone did it together, it would be profitable for all.

1/ Lee, N. and Steadam, I., "Economies of Scale in Bus Transport,"Journal of Transport Economics and Policy, vol. 4, No. 1, 1970. There is no necessary correlationbetween the size of firm and the size of vehicle that the firm operates--smallfirms may operate big buses or large firms may operate small minibuses or even microbuses.

2/ Coase, R.H., "The Problem of Social Cost," Journal of Law and Economics, vol. 3, pp. 1-44, October 1960.

3/ In its essence, Mohring's model did this but with a fixed unit of equipment (size of bus) and an averagevalue of time. These are rather severe limitations. For the bus system is optimized, even given the size of bus, with respect to some average value of time of the average man. Sometimes the average man may wish to spend a lot to save time, or perhaps to get to his destinationwith certainty by a particular time, and he would be willing to pay for an express service. And most people are not average anyway. The suppressionof variety in constructingbus systems for the average is one of the characteristicsof, for example, London's bus system. (In some American cities, express or limited-stopservices are run but usually with standard equipment.)

4/ This argument (and Mohring's) hinge crucially on the fact that the services are not timetabledor scheduled. If services run to sched- ules, the Coase theorem can be easily applied and bargains between runners of adjacent services will ensure the optimality of the result. -4-

1.11 It appears, however, that such effects would be reflected in suitable institutionalarrangements if the costs are low enough. For example, the frequencyof-fct works on a particular route; thus it might pay a broker to act as the solo int4rmadiaryon that route buying the services of the transport agents and contractingthem to potential passengers. The frequency externality would then become internalized.1/ But even if the costs of brokeragewere prohibitivelyhighs it is worthwhile reflectingon the performanceof the competitivemarket in supplying the public with the frequency it wants. Oddly enough one of the bitter complaintsby the operators,not the passengers, against an unregulatedurban bus system is that it gave rise to too many and too frequent buses. 2/ The pirate bus operator who a minute or so before the "regular operator"swooped on the queue at the was the man singled out for excoriation;but he was indeed responding to the demand for a reduction in waiting time. Even in the static equilibriumsense, each competitor would have an incentiveto reduce the waiting and travel time of passengers since quick trips and short waits would command a higher . 3/

1.12 It will also be noted that in other competitivesituations, the same sort of externalityis to be found. For example, consider a number of super- markets serving an area of homogeneousdensity and imagine, a la Mohring, that the supermarkethas to be a fixed size. Then if the populationand the demand doubled, there would be approximatelytwice the number of supermarketsand this would result in a considerablereduction in the access time of customers to the stores. Thus, the marginal valuation of the stores' services would increase.

1.13 Yet there are rarely any suggestionsthat these external economies be used as an argument for subsidizingthe grocery business.4/ Such reluc- tance is partly a consequenceof the belief that external economies are small and in any case probably offset by diseconomies(of congestion,etc.), and partly it could be claimed that the supermarketand grocery business can adjust very well to peoples' desire for quick access by establishmentof higher price small shops, (Ma and Pa shops), specialist shops (such as Seven Eleven) and so on. Over a certair range, varying the size of the super- market by having two small ones rather than one big one--is also a method of reducing access costs at the expense of some economies of scale.

1/ it may be thought that the broker would be a monopolist;however, if there were free entry into brokerage business,this would ensure that he priced competitively. Furthermore,there is always the option of the firms selling directly to passengers and this ensures that the situation will be no worse than the standard competitivecase with externalities.

2/ Such complaintsare to be found in the reports which in the late 1920s and early 1930s preceded the "rationalization"and municipalizationof urban bus services in the West and particularlyin Western Europe, and in countrieswhere minibuses compete with stage buses today.

3/ The great advantage of the competitiveprocess is, however, not to be found in the discussion of static optima. Competitiongives rise to a continuousand urgent seeking of new techniquesand new arrangementsfor giving the public what they are willing to pay for. It is the dynamic process set in operationby the search for profit that gives rise to the singular success of competition.

4/ And in any case it could be reflectedby the state levying lower taxes on small stores. - 5 -

C. Bus Size, Waiting and Traveling Time

1.14 Why do these argumentsnot apply to public bus transport? 1/ It is worthwhiledoing a calculationof the costs of reducing the size of buses from the present standard (circa 40 to 60 seats) and either increasing frequency or reducing running time or both. These measures will reduce the passenger'sinput of his own time and in the latter two cases, there will also be an increase in the utilizationof the equipment. In order to make the comparisonsimple--and, incidentally, to load the argument against the smaller bus--I shall assume that the question is one of replacing one 40-seat stage bus by two minibuses such that the frequencyof service is exactly the same. Thus we can imagine that the minibus picks up at every other stage bus stop--so the stage bus stops twice as frequentlyas the minibus. 2/ A fdcther limitation is that we imagine the service as a typical commuter collector (or delivery) service; commutersare collectedalong the route and delivered to the central business district. The reverse happens in the evening. 3/

1.15 We now need to know somethingabout bus and minibus costs. The evidence of this is rather messy and has been reviewed in Annexes 1 and 2. The broad consensus of this informationis that the capital costs diminish more rapidly than the number of seats. Probably operating costs, other than labor, decrease proportionatelywith seating capacity. Labor input declines less than proportionately;but in order to load the case against the minibus we shall assume that per bus-hour the same quantity of labor is required whether it be a big or a little bus. Thus, the fraction of total cost per hour attributableto labor is larger in the minibus than in the big bus. From one country to another there is a wide variation in the fraction of total costs (excludingsubsidies) allocated to labor--fromas low as 20% in some developing countriesto about 50% in the United States. 4/ For our illustra- tion we take 50%.

1.16 Thus suppose that the cost of a big bus per hour in operation is $20, of which $10 is paid in wages, the cost of an hour's operationof a

1/ The increased congestioncosts seem even more apposite to increased numbers of buses than in the case of stops.

2/ It is convenient to assume that the route is long so that there are a larger number of stops and so that end effects are ignored.

3/ For normal stage bus traffic, this is a point against the minibus service since it does not cater for all intermediatestops. Thus, one could not use the two-minibussystem to go from A to B. Such rigidity is in the model and would not be characteristicof a free competitive system; there is likely to be a variant if there is a sufficientrevenue to be gained from it.

4/ In the United States, "transportationcosts" are about 57% of total bus operating costs--and81% of transportationcosts are attributableto wages. See Table 22 of Characteristicsof Urban TransportationSystems, UMTA, 1974. -6- minibus will be $15, of which $10 is wages. But the minibus operationwill involve fewer stops--approximatelyhalf the number--sothe overall speed of the minibus will be greater than that of the big bus. We may take the speed of a bus as being approximatelyone-half to three-fifthsof the speed of automobile nonstoppingtraffic on city streets. Again, leaning against the minibus, let us assume the speed as three-fifths. By cutting out half the stops, the minibus should attain a speed of four-fifthsthe speed of nonstop traffic. Thus fQr the same number of kilometers as the big bus runs in an hour, one minibus will cost $11.25, and thus to replace one big bus by the two minibuses will cost $22.50--apremium of $2.50 or about 12.5X.

1.17 The question now arises whether this premium would be paid by people for the speedier trip. Again, leaning against the minibus, let us suppose that the average load factors are the same and equal to 50%--so that there are on the average 20 people on the bus. Valuing their time at only $1 an hour, approximatelyone-tenth of the wage costs of the bus, one would find that the 15-minute savings on what was previouslyan hour's ride would give a total time-savingvaluation of $5--i.e.,twice the additionalcost of the minibus.

1.18 It is clear that, if everyone had this same value of time in the bus of $1 an hour, the private enterprisesystem would choose the minibus system rather than the big bus arrangement. Indeed, a value as low as $0.51 would be enough to favor the minibus on this strict comparison.

1.19 We might also briefly explore an alternativeconfiguration and ask whether it would not be better to increase the frequencyof service and leave the running time unchanged. Then the minibuses behave like stage buses--but with one important proviso. Since they are picking up (and discharging)on the average only half the number of passengers per stop, the likelihood that there is no one at the stop will be much larger for the minibus than for the big bus; thus it will make fewer stops. I/ However, let us again ignore this effect with the knowledge that we are biasing the results against the minibus. The total hourly cost of two minibuses is $30, i.e., $10 more than the stage bus system. Suppose that the bus operates a 20-minute frequency and the average passenger is carried for 20 minutes in the bus. Assume that the average waiting time is one-half the , i.e., ten minutes. The minibus will reduce the headway to ten minutes and the waiting time to five minutes on the average. The bus will carry on the average 60 passengersper hour and, under the minibus system, the savings will be five minutes waiting time per passenger--orfive hours waiting time per hour of big bus operating time (per two hours of minibus time). Valuing waiting time at three times ($3 per hour) the value of time in the bus, one gets a minibus saving of $15. And this exceeds the additionalcost of $10. It is clear, provided that the headways with the big bus exceed a little less than 14 minutes (i.e., waiting time averages a shade less than 7 minutes) it will pay to use the minibuses.

1/ If the number of arrivals at the stop are distributedas a Poisson with means m (for mini) and 2m (for bigQus), then the reduction in the probabilityof making a stop is em - e m. Thus if m = 1, the proba- bility of the mini not stopping is 0.368 and the probability of the big bus not stopping is 0.135 (a ratio of almost one third). With m = 2, the probabilitiesare 0.135 and 0.018 respectively. 1.20 These criteria are much affected by the value ascribed to time. I have taken a low value of $1 for riding and $3 for waiting. If we take figures which are nearer to reality in the United States--suchas $2 for riding and $6 waiting, then the critical headway falls to about 6.7 minutes and average waiting time to 3.35 minutes. Thus, if the average passengerwaits more than about 3 1/2 minutes, it pays to switch to a minibus arrangement. Since the minibus arrangementwould be more attractive than the stage bus, then cet par, the number of passengerswould increase and frequencyalso would increase.

1.21 There are obviouslymany other ways to serve these passengers. But as a general rule, the minibus allows more flexibilitythan the large bus. This is obvious for the lightly trafficked routes,especially during off-peak hours when load factors are always far below the capacity even of a minibus (e.g., an average of, say, 5 or 6 passengersper 20-seater or 40-seater bus). But it is important to note that the time savings of the minibus are very important for rush-hour traffic. Thus, if the vehicles are full to their (seating)capacity, on the average, the time saving of the small bus is double that which was calculatedabove (or alternativelythe critical average waiting time falls to 3.35/2 or 1.67 minutes).

D. A Simple Model of Bus Size

1.22 The numerical results suggest that one can develop a simple model of the size of bus and the effect on frequencyand waiting time. Suppose that labor costs per bus hour are the same whatever the size of the bus and whatever the load carried. The bus stopping costs are the same per passenger. The other (non-labor)costs of a bus are proportionalto the capacity of the bus, and taking the load factor as fixed, proportional to the number of passengers.l/

1.23 Now define a "standard bus" which carries a given number of p passengersper hour. (If, for example, each passengerjourney took 20 minutes, the average load on the bus would be p/3 passengers--andthis is assumed to be well below the capacity of the bus.) Let the parameter As be used to indicate the size of bus so that the number of seats is proportionalto X- and for the standard bus )> = 1. If the standard bus is used, the headway gives an average waiting time per passengertrip of v hours and each hour of waiting time is assumed to be valued at %f w, where w is the wage cost per hour of the bus, ( ' >0). 2/

1.24 Consider halving the size of bus, compared with the standard, so that -- = 0.5, then the headway will be halved and the waiting time will, on the average, also fall to v/2. The costs of the smaller buses per passenger will be larger but, to some extent this will be offset by the reductionin waiting time.

1/ In Pnnex 2 it is shown that these assumptions are approximately correct, excfpt that (a) the wage costs decline significantly with the size of bus, and (b) the smaller buses have larger load factors.

2/ The headway is the time interval between buses; thus if h is headway v = h/2. -8-

1.25 One can find the optimum size of bus by constructingthe (continuous) total cost function for p passengers. But since we are interested only in those elements that vary with size of bus, we can be cavalier-likeabout non-wage bus costs and the value of passengertime in the bus. Wage costs for p passengers are equal to the wage cost (I, "b) per bus multiplied by the number of buses per hour. This number is equal to the reciprocal of the headway (with > -1) which is in turn equal to twice the average waiting time. Thus the wage cost term is w/(2 v X ). Passenger waiting costs are the product of number of passengers (p), the average waiting time with the standard bus (v), the index of bus size (,A ) and the value of time ( 'f Thus costs are:

C( w + pvw + non-wage + passenger bus cost journey time costs

-2 dC( ) _-w and d 2v + pv./'w - 0 for minimum costs.

Solving the equation:

_W-2 2, -w~X + 2pv. v - 0, subject to the condition w > 0, we get - 1 /v y2p.' where the positive branch of the root is the only one of interest. 1/

1.26 This formula can be interpretedquite simply:

For a given number of passengersper hour make the size of the bus inversely proportionalto the average waiting time when using a standard bus.

1/ This is a transportationversion of the ubiquitous square root law of inventoriesor batch size. See Herbert Mohring TransportationEconomics, Ballinger 1976, p. 145 et seq. -9-

1.27 And if one is willing to assume that the value ofA/is about 0.5 (and there is evidence that this is sometimes an appropriatevalue), the formula becomes even more simple: 1/ A= l/v F7 To illustratethis: if the average waiting time is 6 minutes (0.1 hours), and if the bus carries 100 passengersan hour, then the standard bus is appropriatesince

A= 1/0.1 = 1 the standard bus is just right.

1.28 It will be noted that for a given average waiting time the larger number of passengers per bus hour the smaller the optimum size of bus. A larger number of passengersper bus hour can be due to either (a) an increase in load factors (which it will be recalled are assumed the same for all sizes of bus) or, (b) a reductionin the average length of journey. The latter point is intuitivelysensible: small buses should be used for short journeys (such as feeder services). The first propositionis less clear: it implies that if there are a larger number of passengers,that is to say the buses are fuller, the bus size should be smaller. But, other things being equal, increase in the load factor simply increasesaggregate waiting time; and a reduction in bus size reducesactual average waiting time. To avoid misunder- standing,it must be recalled that the load factor is always assumed fixed and given exogenouslywhen we examine the optimum size of bus.

1.29 The most unrealisticaspect of this formula is the assumption that wage costs do not rise with the size of bus. Wage costs of the small bus are likely to be far less than those of the large bus.

1/ But such simplicitymay be misleading. Care must be exercised in using this relationship. For example, it is illegitimateto argue that as p-+ 0, A -- 0o so that one would have an infinitely large bus for no passengers! The error arises because v cannot be taken as given and independentof p. For low values of p, the minimum standardvehicle (passengercar) comes into its own and v increases. (And one may add that the assumptionof random arrivals of passengersbecomes more and more tenuous.) Thus if the number of passengersan hour falls to 5 or even 10, one is in the taxi business and responding to demand. - 10 -

1.30 There are two separate influences in labor costs: first, the pecuniary effect of increases in wage rates and associated restrictivepractices, and secondly,the fact that, even under free contractualconditions, bus entrepre- neurs find it profitable to increase their manning scales to include fare collectorsor conductors.1/

1.31 The driver of a minibus commands a lower wage than the driver of the big bus. While there is undoubtedlyless "skill" and training involved in driving a small bus, the main explanationis the virtually ubiquitousunioni- zation of drivers of large buses. This gives rise to much higher wage payments (e.g., more than double in the case of the Washingtonmetropolitan area), but also increaseswage costs by various restrictivepractices such as banning part-time workers, restrictingschedules, slow working, etc. 2/ Unionization is virtually always associated with municipal or state ownership or private (regulated) monopoly. It is conceivable that small buses, if used on a large scale, would become unionized. But the evidence suggests that if the munici- pality or state promotes unions, only weak coalitions would emerge (as in taxis in Washington,D.C.). -

1.32 There are many examples of the use of fare collectors in the western developed countries--particularlywhere fare varies with distance. In London attempts to operatelarge double-deckerbuses with only one man were a failure-- the delays were too long and many of those buses which were designed for one-man operationnow carry a fare collector. In developing countries at least one fare collector is the rule. Bus companiesoften find it profitable to employ two or more fare collectors.3/

1.33 It is thereforenecessary to include manning rates, m (a), as a function of the size of bus. This variable m (;).) can take into account the fact that fare collectorscost less than drivers, and that drivers of large buses cost more than those of small buses. The optimizationthen proceeds by minimizing

C(Q )= w(3A) +,A pv^w + (other terms independentofi ) which gives the equation

2 2 , (2 pvO) A +(dm5 mQ(t) = 0

1/ Or alternativelyoff-vehicle fare collection (and policing) costs will be incurred--suchas those incurred in certain European cities.

2/ Apologists for unions may argue that such high wages are merely rents (of exploitation?)and are not true costs in the social sense. However, it is clear that wage rates well above the supply price of labor have very severe social costs in the form of wasteful substitutionof capital for labor, the persistenteof unemploymentamong the poor, the exclusion of enterprise, and different techniques and so on. I suspect that the difference in financialcosts per unit of output understates the difference in social costs.

3/ This is the case, for example, in Bangkok where two fare collectors per bus are employed. - 11 -

1.34 This can be solved only by specifying appropriatemanning scales which maximize the profits of the organization. It is not unreasonableto suppose that, over the relevant range, the manning scales are linear and increase with the size, i.e.,

m( X ) = a + b a,1 a > 0 and a + b =1

Inserting this linear relationshipwe obtain:

2 2 (2pv X) -a 0 and so

or with

AL = 0.5, I = a/v /v

1.35 This is similar to the formula derived above, but now the term a appears in the numerator. This term is less than unity since, for the standard bus, we have a standarizedmanning scale of m = 1 in the formula above. Thus, a + b - 1 and if b is strictly positive then 0 a L 1. Thus in Figure 1.1 we have plotted the manning equation with a 1/3, b 2/3 when the standard bus has 40 seats.

Figure 1.1: MANNING SCALES for m 0.33 + 0.67 A

m l

a a______Iseats 20 40 60 80 100 )k= 0.5 =l 1 = 1.5k= 2 A= 2.5

1.36 For the standard bus m = 1, and the minibus with 20 seats has a manning scale of m = 0.67. The 80-seater has a manning level of 1.67 and so on. - 12 -

1.37 In using this formula, therefore,we observe that the effect of wage-costsvariation is to depress the size of bus by the multiplicative factor a. The higher the value of a the greater the manning economiesof scale. Thus if a - 1/3, M. - 0.5, v - 0.10 and p = 100.

i - 1/3 i.e., the size of bus should be one-third the size of the standard bus. (This, of course, may be beyond the applicable range and one may choose the smallest size of bus with the lowest wage cost.)

1.38 It is interestingto note that these are the approximatevalues which one obtained from Washington,D.C. if the taxi, the medium-sizedbus (22-seat school bus) and the large metro bus are compared. 1/ Assuming a waiting time as little as 3 minutes for an average number of 144 passengers an hour during peak periods, this suggests that:

A = 0.33 (0.05;' '44) and thusA= 0.55, the size of bus should be approximatelyhalf the present 55 seaters. A smaller 30-seat bus would be, on the average, more appropriate.

1.39 For Manila we can calculate the implied values of a and b from the labor costs of jeepneys and buses (see Annex 2, Table 1). The appropriate equation is m = 0.148 + 0.852 A where A. is unity (and so m = 1) or the standard 58-seat bus. The value of a implies that small buses are appropriate in Manila. With v = 4/60 (i.e., 4 minutes) and p = 144, we find A = 0.185, or that the best bus is about 11 seats (0.185 x 58). This is quite near to the standard size. However, usingAM = 0.5 to refer the value of waiting time to the wage costs of the standard 58-seat bus is clearly over- stating the valuation of waiting time. A more appropriatevalue would be,6. = 0.2. This would imply that we multiply (0.185 x 58) by 45/2 to find the value of A -- giving#R = 17 seats. It is perhaps relevant to note that the small jeepneys have been more successful than the large buses--but this needs investigationand is explored next.

1.40 Although these are clearly rather rough approximationsto the value of a, they do show the critical role played by economies of scale in labor costs in the bus business. It would be convenient if one could set sensible limits to the values of a that one might encounter in practical situations. When a.is near to zero, this implies that the economiesof scale are rather small, whereas when a is near to unity the economies of scale are very large indeed. We have seen in the calculation for Manila that a small value of a is distinctly possible and one might suppose that the lower limit is around 0.10.

1/ These are $9,000 a year for the 5-seat taxi costs, $13,000 or so for the 22 seater and $23,000 for the metro driver (including fringe benefits at cost)--all 1975 values. The fitted equation to the two extremes is m = 0.33 + 0.67A. There is a strong union effect in the very high wages of metro drivers and in the higher manning requirements. This is, however, a very crude comparison and is a poor substitute for an adequate analysis of wage costs and manning scales. - 13 -

On the other hand it is clear that the upper limit of a is not near to unity. Probably an upper bound of about one-half is most appropriate. This implies that a double-deckerbus of 80 seats would require wage payments of 50% more than the single-decker40-seat bus. (Similarly,the minibus of 20 seats would have labor costs of three-quartersof those of the single decker 40-seat bus and half more of the double decker.)

1.41 To illustratethe applicationof the formula,with a value of a equal to one half and one third (the "D.C." value), one can take the data collected by Mohring for Minneapolis. The average waiting times, using the standard bus during the peak and off-peak were 4.65 minutes and 7.0 minutes respectively;the averagebus occupancy was 30 minutes during the peak and 10 minutes during the off-peak. Assuming that the average journey was 20 minutes, this gives a total number of passengersper hour of 90 and 30 respectively.1/ With a = 0.5 we obtain:

* peak = 0.68

off-peak = 0.70 and with a = 0.33

- peak = 0.45

off-peak = 0.47

1.42 These results suggest that the size of bus in Minneapolisshould be between half and two-thirdsthe size of the present vehicle. Instead of a 55-seaterbus, it would be best to run a 27- to 33-seat bus. Oddly enough the size of bus which would be best for the peak periods appears to be somewhat smaller than that for the off-peak periods. This illustratesthat there are roughly three times the number of passengers (90) compared with the off-peak (30) who are waiting 4.65 minutes each. This large number of passengerswould benefit in aggregatemore from the reductionin bus size and increase in frequencythan would the 30 passengersduring the off-peak period.

1.43 One of the attractive features of this formulationis that the size of bus is largely independentof the price of labor; it applies to high-wage and low-wage economies. Wage costs are counterbalancedby the costs of waiting. In the standardizedformulation, however, it will be noted that the wage cost is that of the standard bus and not necessarilythe wage rate of the driver. This can be taken into account, as we did in the case of Manila above, by adjusting the value of . And also, since it is likely that the marginal costs of labor (especiallyunionized labor) rise steeplyduring the peak periods, it may be necessary to take that effect into account in deciding on the best size of vehicle. Such refinements,however, seem hardly worth pursuing in a general formulation.

1/ Since people have to wait so long during the off-peak, one might expect them to travel on the average further than journeys during the peak period; but this is probably offset by the higher off-peak speeds. - 14 -

1.44 It is important to note the status of this formula and its applica- tion. In my view it would be quite wrong to plan a transport system on the basis of such a formula, and indeed any other formula such as those suggested by Mohring and Nash. To be simple, any formula has to deal only with averages and cannot take into account the vast variation around those averages. In particularit is clearly true that differentpeople have enormouslydifferent values of time--oftenthe same people have very differentvalues of time in different circumstances. Equally there is a wide variety of transportroutes which may best be served by a great variety of vehicles and organizational forms. The formulas give one a meat-axe critique of existing systems or proposed dirigismesystems. Neither one nor the other should be unthinkingly imposed on sufferinghumanity.

1.45 The second point to emerge from the evidence is that minibuses manage to get a much larger load factor than stage buses. 1/ The assumption of the same load factor is clearly discreditedby the evidence. One would, of course, expect such a result. But it may be argued that in certain cases the high load factors on minibuses are due to the restrictionson the supply which are more onerous than those on big buses. However, the same phenomenamay be observed when there is little or no restrictionon minibus operation (as, for example, in the early days in Kong and Bangkok). It may well be that this is mainly a consequenceof the greater incentivespresent in the small firm, often an owner or renter-driver,which characterizesthe minibus trade. If the small firm were to be allowed into the larger bus business, it is conceivablethat the load factors would also be higher--butthere is little evidence on this hypothesis.2/

1.46 The third point is that, with the normal mixture of origins and destinations,there is likely to be much less transferringnecessary with the small bus system. It is more likely that a passengerwill find one bus that goes all the way. This is an importantadvantage of the minibus since it is well known that a transfer is a considerabledeterrent not only with transfer-timewaiting costs equal to three times the on-bus time costs but also because it adds an additional element of discomfortand uncertainty.

1/ For example, in Kuala Lumpur, the peak load factors were 0.67 for the bus and 1.07 for the minibus, and in Belfast the load factor of the black taxis was two-and-a-halftimes that of the bus.

2/ Minibuses are often accused of "creaming" the traffic--whichmeans roughly that they serve the services where passengerdemand is highest at profitable fares. They are said not to serve the "sociallynecessary" loss-makingroutes. This may be true--but the fault lies with regulated fares which prevent a profitableminibus service. In practice, however, minibuses often provide the only service in certain areas (e.g., in the sois in Bangkok) where no stage buses will venture. Similarly,in Istanbul,the minibuses are given routes which have been rejected by the municipal bus authority. - 15 -

1.47 The last point to consider is the effect on road congestion costs. Superficially, it may seem that the replacement of one large bus by two or more small ones during the peak hours would be bound to cause additional congestion costs. Yet reflection will show that the case is clearly not proven. It is normal to count a bus as equal to three cars or two medium-sized trucks in order to get passenger car units (pcu's) for congestion analysis. 1/ In terms of road occupancy a minibus will lie somewhere between a large automobile and a medium-sized truck. If the same total number of stops are made, however, as in the first model discussed above, then it seems that there can be little effect on congestion whether we run two minibuses or one big bus. In fact, the minibuses would probably result in a slight reduction in congestion since they are apt to be more lively in acceleration than the standard bus. 2/ Probably the most important factor in the minibus favor is the high utilization and load factor. In Istanbul, Sanli (1977) concluded that, per passenger, the minibus (10 seats) is twice as efficient in terms of road occupancy compared with the standard large bus. But much will depend on the types of operation, customs and laws in the use of the highway. Many other configurations can be considered--but at present the evidence suggests that, for the same quantity and quality of service, the congestion effects are small. If the quality of the service is improved for a given price, the minibus service is likely to attract car drivers and passengers and this substitution will reduce congestion. We conclude that the congestion effects are likely to be small and unlikely to offset the direct effects of the substitution of small for large buses. 3/ But this remains a weak conclusion. Congestion and minibuses remain unfinished business.

E. Regulation and the Consequences

1.48 Before reviewing the experience of small buses in the real world, it is necessary to set out some characteristics of the typical situations, both regulatory and economic, of urban bus transport. To illuminate these situations, Figure 1.2 sets out the normal two-service market situation where there is competition between the bus and the minibus. In Figure 1.2a the demand curve D (M = 0) shows the demand for bus trips when there are no minibuses avaiPable. When minibuses are available at price p however, the demand curve for bus passenger trips is shown as D (p ). The demand for minibus passenger trips D (pb) depends on the price oT the bus passenger trip and is shown in Figure l.Tb. 4/

1/ See Smeed & Wardrop (1964). 2/ Thus in , the public light buses are rated at 1.5 cars for acceleration from a stop light. 3/ Noise and pollution are two other externalities that are normally considered. I suspect that the minibus would gain on quietude but lose on pollution (particularly if gasoline rather than diesel engines were used). But such a comparison would take us far afield. On noise, see National Research Council, Noise Abatement, Policy Alternatives for Transportation, National Academy of Sciences, Washington, D.C., 1977 (Chapter 5). 4/ All the ensuing analysis refers to passenger trips--but occasionally, whenever the meaning is obvious, they are referred to as bus or minibus trips. - 16 -

Figure 1.2: COMPARISONOF TRIPS: BUSES/MINIBUSES

$ 1.2a: BUSES $ 1.2b: MINIBUSES

pn PbSX Dm(Pb)

No. of No. of I_ passenger - passenger 0 B2 B1 trips 0 M trips

1.49 Now we can review first the circumstancesof cities where there are no minibuses and only stage buses of the traditional size (over 40 seats). The first archetype is.where there is a minibus ban; the authorities prescribe that only full-sized buses are allowed. This corresponds to the authorities specifying a supply curve of minibus passenger trips that is coincident with the vertical axis; it does not matter what their costs are, no minibus service will be supplied (M1=O). Let us suppose that in the big bus business the supply curve is horizontal and that the supply curve is coincident with the price p . (Alternatively, one may suppose that there is a regulated fare of Pb which is in excess of marginal costs of bus passenger trips, and that licensure of bus capacity restricts the quantity of trips.) Thus the number of bus passenger trips is B1 at price Pb-

1.50 A second archetype is indistinguishablefrom the first; this is where the costs of minibus passenger trips exceed the regulated maximum fare which is imposed for all bus journeys,whether by minibus or stage bus. Thus suppose that the marginal and average costs of minibus trips are p . Then the fare regulation at pb will ensure that no minibus businesses are set up-- although people would be willing to pay more than the costs for the services of the minibus. In fact, a particular variant of this model seems to be common. This is where the bus operation is a municipalized or franchised monopoly. In general, the monopoly will find it more costly to operate minibuses than stage buses (as Nash and Mohring assume)--partly because it has unionized labor and partly because of the inherent rigidities of scheduling and operation in a - 17 - large firm or bureaucracy. Minibuses, although providing a superior service, could only be operated at a loss at the existing regulated fare. 11 This franchised monopoly case is entirely different from the situation where there are conditions of more or less free entry (or at least no effective proscrip- tions on entry) into the minibus industry. Then, although the fares may have the same regulated maximum, the likelihood is that small operators will find it possible to earn at least normal profits at those fares which are thought to be only just compensatory or normal for the large franchised monopoly stage bus company. (This seems to be a good approximation to reality in the cities of many developing countries, e.g., Istanbul, Manila, etc.)

1.51 The third case to consider is that where minibus fares are not effectively regulated, and one may conveniently suppose that there is no really effective ban on entry so that there is no monopoly control. Then, if the authorities control the stage bus fares nicely at cost (equal to p in Figure 1.2a), and if the costs of producing a minibus trip are constant at Pm, there will be M minibus trips competing with the B2 bus trips. There are, as we shall see, many variants of this scenario. For example, in Malaysia (Kuala Lumpur), an effective quantity constraint is placed on the minibuses, and although their fares are regulated they are considerably above the costs per passenger trip; yet the passenger demand is such that there is considerable overcrowding of the minibuses.

1.52 To summarize these propositions, therefore, one would not expect to find any substantial number of minibuses where there is a monopoly, either municipalized or privately controlled, in the supply of bus services. The high costs to the monopoly of small buses, together with the regulated fare level, will ensure that minibuses have a small role, if any, in urban trans- . The monopoly will find that in the absence of a fare differential the only way it can recoup the additional costs of the small bus is through a larger number of passengers due to the shorter waiting time. This demand effect is unlikely to outweigh the additional cost, hence the widely held view in municipalized monopolies, enunciated by Nash, that the larger the bus the better. If, on the other hand, the monopoly applies only to large bus compa- nies and there is at least the p2oiility of entry, much will turn on the conditions and particularly the regulation of fares that are imposed. Even in the most unfavorable circumstance, where minibus fares are regulated at the level of bus fares, one would expect to see substantial entry of minibuses--the extent would be the result of regulation rather than by profit and loss. And such a result would be true a fortiori when the regulated minibus fare exceeds the bus fare. Finally, under conditions of free entry and free fare setting one would expect that minibuses in small firm owner-driver form would become a dominant mode, probably accounting for more than half of public transport trips. Then the individual operators will be able to increase passenger fares for their superior, more frequent service.

1/ In some cases, for example, in certain limited, lightly trafficked routes in London, London Transport put on minibus service at higher fares; but these were a minor exception. - 18 -

1.53 However, as we argued above, with numerous suppliers of bus service, there will still be a case for further increasingfrequencies through using even smaller buses. This externalitywill still be present, and there is a case for introducinghigher operating taxes (or smaller subsidies)per seat on the larger buses. It is also worthwhile reflecting that in a mixture of minibuses and buses, all buses whatever their size will enjoy the advantage of the increased frequencies brought about by a preponderance of minibuses.

F. Case Studies

1.54 The history of minibuses is now briefly reviewed to see whether there is any revealed evidence of their efficiency. 1/ Two caveats must be borne in mind. First, in general one can observe only surviving minibus systems; those that failed or were eliminated in the competitive process do not figure substantially in this review. It is worth emphasizing that there appear t:obe very few cases where the minibuses or jitneys have been eliminated in a free competitive system. 2/ Secondly, one must always beware of the fact that one may identify so-called advantages of a minibus system whereas what one is really observing is the advantage of the competitive situation. 3/

1.5:, In two countries, Argentina (Buenos Aires) and Honduras (Tegucigalpa), there has been freedom for the operator to choose his size of vehicle. In other respects, however, there is control of fares and, in the case of Buenos Aires, stringent control of entry. In Tegucigalpa, the predominant vehicle is the Volkswagen 10- to 14-seater bus. In recent years the increase in the tax on gasoline has caused some decline in the popularity of the Volkswagen gasoline bus and some shift to the somewhat larger diesel buses--but the 10-to-14 seater remains the modal model. It will be noted that the fare is the same whether one by the more frequent small bus or by the less frequent: large bus. In Buenos Aires, in a rather high wage country compared with Honduras, the average size of bus or "collectivo" is in the region of 21 seats. This is considerably less than the standard size bus in places as diverse as Calcutta and London. Buenos Aires originally had a nationalized monopoly public bus company and characteristically ran large buses. But the business was denationalized and taken over by private operators, who reduced the size of bus considerably. Now the operation is very efficient and the zonal.flat rate fares are only $0.15-0.23 for this large city.

1/ The evidence in this section comes from many sources. Those sources that are published will be found in the references. Some information, however, is based on unpublished material and some statements are derived from direct inquiry or observation of the systems in practice.

2/ As far as I am aware, only the jitneys of Hilo, Hawaii were the victims of competition from other modes.

3/ This may, but need not necessarily be associated with predominantly small firms. But in a free entry situation, it is possible for quite large fLrms to exist side by side with a large number of small firms; this can be seen, for example, in the trucking business in the U.K. I cannot follow Nash (1978) who asserts that there is "an inevitable (high) degree of monopoly power in public passenger transport" unless he supposes that a competitive system is "politically" unacceptable or impossible. (op cit., page 84.) - 19 -

1.56 Incidentally,this disposes to some extent the argument that it is only the ownership structure (owner-driver,etc.) that makes the m'nibuses successfuland efficient. The small firms in Tegucigalpacould have chosen the larger vehicles; indeed some did, but most did not. In effect, the free market choice shows that there is a wide variety of equipment and capital structures that correspond to the variety of entrepreneurialability. Some operators are efficient at running large vehicles, but most appear to be best adapted to the small ones. 1/

1.57 Another sort of comparison can be made in cities where there are large bus and minibus services running side by side. In interpretingthese situations,one must bear in mind that both big bus and minibus firms are often regulatedwith respect to the fare, the vehicle size and sometimes even the area or route. The effectivenessof regulationvaries, but generally it is applied rather more rigorouslyto the large firms with stage buses than to the small minibus firms. It is more difficult to regulate a large number of small firms. Governmentsparadoxically try to encourage collusion among these small firms so that they can be better controlled. Since one of our tests is the profitabilityof the services,it may be that the net revenue is much affected by the ability of the minibus owners to bend the rules. Some idea of the extent of this evasion of regulationwill be given in each case.

1.58 Manila (Philip2ines)is one striking example of the simultaneous running of jeepneys, (jeeps converted, usually quite estheticallyfor carrying about 14 seated passengers)and standard buses. The jeepneys are often driver-ownedor sometimesrented by the day or week; the buses are privately owned by largish firms. The fares are the same for jeepneys and for buses. Jeepneys, like buses, are nominally restricted to certain routes, but it is doubtful if the regulation is rigorously enforced. Jeepneys are very frequent and tend to stop on demand and will detour to deliver passengers to the door. On the other hand, the average length of journey by jeepney is somewhat shorter than that by bus. The evidence shows that although the bus companies had overall deficits at the regulated fare, the jeepney owners were making substantialprofits. In part, this is because bus drivers, supposedlymore skilled, were paid higher wages than jeepney drivers, but on the other hand jeepneys appeared to be better maintained than the buses.

1.59 This evidence from Manila is, of course, not inconsistentwith the view that there are substantialdiseconomies generated by the ownership and (lack of) incentivestructure in large companies (bus companiesmust own at least three vehicles to become licensed).

1/ To avoid misunderstanding,this does not necessarilymean that, granted a large firm is in operation,either by decree or protection,it would not usually be wise for it to operate large buses. If fares are con- trolled and the firm is not induced to take into account the full value of waiting time, it may well be consideredbest to run large vehicles. - 20 -

1.60 Hong Kong has privately-ownedlarge franchisedbus companies (the main one being the Kowloon Bus Company) operatingmainly double-deckers(73 seats). The minibuses, called public light buses (PLBs), have 14 seats and are owned by small firms, (74% were one-vehiclefirms.) 1/ The PLBs were developedillegally and in their initial years were reported to have been making fabulous profits--figuresof 300% on capital were not uncommon. In 1969, PLBs were legalizedbut they were not controlledwith respect to fare or to route--exceptthat they were prohibited from certain roads where it was thought that they would add unduly to congestion. (A PLB is counted as 1.5 private cars for congestionpurposes.) The standard fare of the big bus is HK$0.30 whereas the PLBs tend to charge about HK$0.40-0.50and levy an addi- tional charge to make the fare HK$0.75-1.00or so during the peak period. The frequencyof the PLBs is on the average about 8 times the bus frequency, giving average waiting times of 4.7 minutes and 33 seconds on comparable routes and their speeds are on the average 15% higher. The nearly 4,000 PLBs carry one-third of all passenger trips by public transport.

1.61 Financialdata are not easy to come by, but what evidence is avail- able suggests that the rate of return on capital for a PLB is still more than 100% per annum and considerablyin excess of the rate of return to the stage bus companies (which remains at about 20%).

1.62 One interestingrecent developmentis the likely effects of the opening of the Metro (rail underground)system. In order to pay its way, the Metro fare is calculated to be about HK$2.00, and it is suggested that to induce passengers to use rail rather than bus, the stage bus fare should be increased. The stage bus operatorsobjected that they now earn reasonable profits at the present HK$0.30 fare and did not want an increase!2/ It is not difficult to see why.

1.63 Chiengmai, a city of about 100,000 inhabitantsin northern Thailand, provides an interestingillustration of virtually free competitionbetween ordinary stage buses and minibuses. 3/ The typical minibus is a 1,000 cc or 1,500 cc utility offering about 12 to 14 seats and costing about $3,750. The typical 30-seat bus costs $22,000 to $25,000. With some importantexcep- tions, the minibuses did not operate on fixed routes but respondedto the demands of passengers,whereas the buses all adopted fixed routes. The normal flat rate fares were $0.05 by bus and $0.10 by minibus. 4/ But on costly routes during night hours the minibus operators increase fares above the $0.10 level. In spite of the higher fares more than 90% of public transport trips

1/ See also Annex 2, Table 2.

2/ Hong Kong Standard,June 1, 1978.

3/ See P.R. Fouracre and D.A.C. Maunder, "Public Transport in Chiengmai, Thailand,"TRRL SupplementaryReport No. 285, 1977, Transport and Road Research Laboratory,Crowthorne, Berks, England (1977).

4/ Until 1974 the fare by minibus was only $0.05. - 21 -

are by minibus. The bus company which is described by Fouracre and Maunder as doing "moderatelywell in comparisonwith other bus undertakingsin terms of physical measures" is not a profitable concern and is "evidently in decline." "It is difficult to see how it manages to survive" (op cit. p. 6). And it is pointed out that the bus company does not increase its fares because it would lose a substantialamount of traffic to the minibuses. The minibuses,on the other hand, are modestly profitable,as one would expect from the more or less free entry conditions,and provide a valuable source of productiveemployment (op. cit., p. 23) at competitivelevels of wages. The main conclusion to be drawn from the experience of Chiengmai is that under free entry conditions in medium-size cities the passengers prefer the convenienceand time-savingof minibuses even at double the fare of stage buses. One feature of the Chiengmai minibuses is difficult to explain--thatis, the very low load factor, 0.072 compared with a bus load factor of 0.21. 1/ This remains an anomaly which must be viewed with some skepticism until resolved satisfactorily.

1.64 Kuala Lumpur introduceda restricted number of minibuses (see also Annex 2, Table 2) in 1975 in preparation for the introductionof an Area Road Pricing Scheme in the expectationthat the better service would lure some car owners out of their vehicles. The number was limited to 400, then raised to 450, mostly one-vehicleowners, compared with about 520 stage buses which were organized in large private companies. The capacity of the minibus was 16 passengers and no standing passengers were allowed, whereas the capacity of the stage bus is about 50. The fare on the minibus was a flat M$0.40, whereas the fare on the bus is graduated according to distance starting at M$0.10 and increasing at M$0.05 a mile--the average fare is about M$0.21. The average load factor on the minis during peak periods was 1.07 (illustratingthe fact that the no-standingwas not enforced) and that the number of minibus licenses was the binding constraint. The buses ran at load factors of 0.67. Naturally, the minibuses took the lion's share of long-distanceroutes with an average length of trip near to 11 km. They accounted for about 35% of total inbound trips and respondmore flexibly to the peak by carrying an extra 3% (i.e., 38% in the peak).

1.65 As one might expect, the minibuses.were alleged to be doing very well indeed, and there is an intense demand for new licenses,whereas the stage buses were, on the average barely earning profits. The stage buses were understandablyreluctant to expand their services unless the minibus competi- tion were contained.

1.66 In a number of cities, the taxi has been used as a fixed route service similar to buses. Sometimes, it has been illegal or extralegal,as in the case of Belfast (Amos 1978). About 1971-72 in the Catholic areas of Belfast, the municipalizedcity bus operationswere severely curtailed,and owner-driversset up an alternativefixed route service by taxi at frequencies much greater than bus, and at fares considerablybelow those charged by the buses. 2/ Subsequently,the decline of police regulation saw the black taxis

1/ These were calculated from Table 20 of Fouracre and Maunder.

2/ Of course, the IRA had the choice of stealing city buses for this service--andit is of interest that they chose to buy old London taxis. - 22 -

spread to protestantareas. The nominal capital costs are, of course, rather small. Little is known about the labor cost, but it is known that the load factor is two or three times the comparablebus load factors of 0.2 to 0.33. The IRA apparently earn substantialprofits from this black taxi service--asdistinct from the city bus undertakingwhich, as so often is the case, is substantiallysubsidized.

1.67 In Istanbul there are three basic modes of public transport: the dolmus (a shared taxi), the minibus (8 to 10 seats), the midibus (13 seats), and the standard large bus, with a few trolley buses. The latter are operated by a publicly-ownedcorporation, Istanbul Electric Tramways and (IETT), whereas dolmus,minis and midis are owned and operated by, very largely, owner-drivers. (We shall refer henceforth to midis as simply minibuses.)1/ The approximatenumber of vehicles licensed to operate in 1976 and data on the number of passengertrips in 1978 are as follows:

ISTANBUL PUBLIC ROAD PASSENGER TRANSPORT

No. Passenger Passenger of Trips Trips Licenses ('000 per day Per vehicle (1976) est. 1978 Per Day

Dolmus 16,000 1,250 78

Minibus 3,500 1,100 310

IETT Large /a Buses 1,070 -- 625 1,450

/a These data refer only to licensed authorizations,and by December 1978, due to breakdowns,etc., the number of IETT buses actually in service was only about 450 or less. Mini- bus trips were on the average longer than either bus or dolmus trips.

Source: Sanli (1977) and author's estimates.

1.68 Entry is restricted into dolmus business by the limitationsof licensing;however, it is clear that in practice there is a substantial unofficialoperation which is the constant bane of the licensed dolmus opera- tors' association. The municipalizedundertaking IETT has first choice on routes and then relinquishesthe least desirable routes to minibus operators who "never say no" and become the approved operators on these inferior routes. (Dolmus are not restrictedwith respect to routes.) Vehicle seating capacities are regulatedby law; thus a dolmus has only up to 7 passengers,and a (strict)

1/ See H. Ibrahim Sanli, Dolmus-MinibusSystem in Istanbul, Istanbul 1977. - 23 - minibus up to 10 seats. Fares are regulatedor formally approved by the authorities. Bus and minibus fares are roughly the same--US$0.10for the first 2 1/2 kilometers and rising rather slowly with distance. At these fares the IETT bus operationcovers only about 35-40% of its cost, whereas, notwith- standing very high taxes on and scarcity prices of minibuses, the minibus operators make at least normal profits on these "inferior"routes to low- income squatter areas (p. 120). 1/ The load factors are very high on minibuses--thenumber of passengersconsiderably exceeding the number of seats during many hours of the day; and on IETT buses, the load factors were also high. However, Sanli concludes (p. 54) 1/ that minibuses have road occupancy efficienciesequal to twice that of the fully loaded bus. Dolmus fares are about double those of bus fares for short distancesand almost four times those for the most distant bus destinations. Nevertheless,with old, sometime pre-World War II American automobiles,operators earn at least normal profits.

1.69 Istanbul aptly illustratesmany of the points of this paper. The municipalizedbus undertakinghas a highly unionized labor force with high wage costs (about twice those of the private sector (Table 1.1)) and has about 17 persons employed per bus in service. The fixed low fares cover only about 30-40% of costs making it difficult-tofinance new buses; so the service declines, bus miles decrease and load factors increase. But the private sector minibuses, despite the difficultiesthat they encounter,can operate profitably--butonly at very high load factors. These are maintainedby restrictionson licenses. And so the final frustrateddemand for cheap public transport is squeezed out to the dolmus and induces the operation of illegal dolmus. But the role of the dolmus in providing frequent fast and "per- sonalized" service gives a superior, although more than twice as expensive, alternativefor those who are willing to pay as well as providinga fringe for the frustratedbus traveler. The arithmetic of the situation is also compel- ling. If IETT increasedbus fares to break-even levels, they would be in excess of the existing dolmus fares for all except a few very long journeys, q.e.d.

1.70 In Bangkok (approximatepopulation 5 million), there are known to be about 4,000 stage buses. Until 1976 these were owned by separate relatively efficient private firms, but followingthe advice of a European consulting firm they were all amalgamatedinto one authority--theBangkok Metropolitan Transit Authority. The fares are controlled by the governmentat 1 baht (about US$0.05) at a flat rate. There are an unknown number of minibuses (the figures commonly used vary between 3,000 and 7,000) operating de facto but not de jure in the metropolitanarea. Minibuses, since they are illegal, are not subject to formal control and little is known about the details of their work. They respond to demand and serve the unpaved side streets (sois) and out-of- the-way suburbs where the stage buses do not venture. They charge the same fare as the stage buses (exceptit is said at certain peak periods of conges- tion or in the middle of the night). But at this fare it is clear that the minibus is making at least sufficientprofits to induce them to stay in

1/ Ibid. - 24 - business,whereas the BMTA is sufferinglarge financialdeficits. 1/ An increase in bus fares, the normal solution for bus deficits,would encourage -he influx of more minibuses--andit is noteworthy that the authoritieshave .e..tly decided to legalize and control the minibuses.

G. A Summary of Empirical Evidence

-.5 71 The ,irst striziingfeature of Table 1.1--whichsummarizes the -a ir.-."a_ :l2., icei t- in the two cities where there are no effective >-4 oGnentry the minibuses have ceptured more than 80% of the market. 5T, B-'-es os the owner-drivers of the collectivos, although regulated on r'' awe,ree to buy bigger vehicles but did not.) The review of the it- v suggests that, were entry controls removed and no increase in Es Y0 biDuses there wouild be considerable gains for the miinibus in the : .-.:E:: es, nne mas con'ecture that manir7uses would capture at least

the toEal passenger kilometers. Secondly tt appei. that the _' j ' .. esGthe role o f minibuses--carrying long-distance trafic instead of . ltn e ls bro-ugin- about by government regulationr and is not an - e-eot i4n the aarhet system. Thirdly, the high pro` s of m-r:-!bus all directly attributable to controls on entry--duly e-;acerec£ Y: at ation of stage buses. The exception is Belfast-but it is a otnet'her the government's writ is observed in the Catholl c areas c f ' '3n-try is controlled by the sectarian groups. Fourthly, with the C. c-.-.Ssezeption of Chiengmai, the load factors of minibuses seem to be .sghser and sometimes very much higher than those of the stage buses. N t c.aan vehicle utilization have been included in the table, but if it t b e listed it would look very similar to the column on load factors. is .orth emphasizing, finally, that all the minibuses operated in these c.ties wi^?ereprimarily owner driven. They were quintessentially small operators. The stage buses were usually either larger companies or a municipalized monopoly.

i.72 It is interestingto reflect on the reasons why the authorities choose ithe large bus. In cities where there is a viable minibus service, there is usually an underlying feeling by the authorities that the minibuses should be suppressed. (The notable exceptions are Kuala Lumpur, Chiengmai, Hong Kong and perhaps Bangkok.) It is clear that municipal authorities have only inadequate incentive to react to the public's demand for a reduction in waiting time. Fares are regulated and passengers are not allowed to bid higher for a shorter wait (except insofar as they use taxis). A change in waiting time will affect London Transport only insofar as it changes the number of passengers traveling at the fixed fare. And if demand is relatively inelastic (because of the costs of parking cars in London), there is little revenue to be lost by increasing waiting time by using the large bus. Passen- ger waiting is only fractionally reflected in the accounts of London Transport.

1/ Before "integration" of the bus companies into BMTA they were on the average making reasonable profits. Since there is no formal provision for financing deficits, BMTA is covering these by raiding its depreciation funds, maintenance operations, etc. Table 1.1: A SUMMARY OF STATUS OF MINIBUS z~~~~~~~~~~~~~~~~~~~~~~~~ Share of Minibus in Total GovernmentRegulations Length of Trip Load Factors Profitability Bus Pas- on Comparedwith Compared to Mini- Stage senger Km Entry Fare Stage Bus Stage Bus bus Bus Belfast 20-25/b None effectiveNone Normal High High/a Unprofitable in relevant areas

Buenos Aires 85 On number Flat rate n.a./d Normal to - agreed medium-high Bangkok 20-30/b Illegal None Not known Normal Normal Unprofitable Chiengmai 90 None None Short Very low/b Normal Unprofitable

Hong Kong 38 On number None Normal Normal Very high Normal L

Istanbul 80/c On number Dolmus: Long for Very high on Above Very large Twice iuinibuses minibuses normal losses stage bus Mini: same as bus Kuala Lumpur 50 On number Flat rate Very long Very high Very high Normal to exceeding low bus Manila 60 On number Flat rate Short Medium-high Normal- Unprofitable same as on high bus side Tegucigalpa 85-90La None None Normal Medium-high Normal Normal

/a Uncertain. /b Highly uncertain. /c Includes dolmus in minibus category. 7d Not applicable. -26-

1.73 A secona reason is that often the big bus is considerably subsidized. This may take various forms. The most harmful is the 80% capital subsidy by the federal government in the United States--capital is then almost a free go.~ 3ut like the free lunch, gifts from the government have very expensive string r.ttached. Other forms of subsidy are low taxes on diesel fuel, bus _-..sfrom import duty, lower license fees, and preferential traffic arr a: ,-at: s

A _aird reason is, I suspect, most important and that is the fasci-

sa-ion with the large '.-.tegrated" or "coordinated" transport system. 1/ '3 -e, lanner. mass-. transit means centralized control and large _}eiih. -d by trbuta4ries of smaller veh.icles. The large vehicle has - .? LoTce costs according to their calculations.

,:.astly.it will be noted that munlcipal transport undertakings are ; J,zed on the e hand, and lack-the incentive to seek profits on - is,-- the s, 2_f4 of somno'Lenceand lethargy. Th-t also high wages a .se of Washiro-on_ DaC.) will Induce hngh capJtaliza'fo4n at the -m_.-:s soSid uinion ranks -will oppose any incursion o£ _ne smaller

Conclusions on Policy

T_a Dolicv implications are complex and depend on mriany co-niderations sconomic issues. In perhaps most cities of the world, the bus .-e is ulnicipalized and substantially subsidized. It is accepted as a 20^-ax4ortatic that all urban road passenger public transport undertakings mus- - sub3idized. In the United States and many other countries, these take of capital subsidies (where the federal government pays for about 80% .:th sosst of the bus) and some underwriting by local authorities of operating Ihe form of the federal subsidy in the United States and many other cou-,t-`,es enacourages high capitalization which gives rise to lower frequencies and not: the higher frequencies which Moh^-.ng suggested as a rationalization for autosidy. Similarly, the unionization of labor, indeed a condition of the federal subsidy in the United States, encourages high wages and so supports the overcapitalization.

1/ Since 1951 I have attempted to understand what is meant by integration and coordination--indeed I wrote a book about it--but the only sensible meaning I could determine is that which simulates private enterprise market solutions. Transport planners believe, however, that they can produce solutions far superior to the market; one's experience in many countries casts considerable doubt on their claims. - 27 -

1.77 The municipal undertaking will have a vested interest in showing that its choice of vehicles was correct. It would be against the grain for it to pursue enthusiastically the provision of minibus services in competition with its normal services. Similarly, the union would be hesitant about trying new practices and would probably be aware of the possibility that such services may be given out to the private sector in the form of shared taxis or the like, which would reduce union membership. The authorities are therefore unlikely to have the courage of the Malaysian or Hong Kong governments and authorize a rapid expansion on such services.

1.78 One alternative method is to identify those routes which are run because they are a "public service" and which incur the maximum losses. These could be hived off. (This was the very successful "way in" for the minibuses of Istanbul.) Thus to hold down the deficit of the municipalized service, the authorities could license any operator who wishes to offer a public service at any fares he chooses. The operator may be required to have suitable insurance and to post maximum fares on his vehicle. Beyond this, no other control would be imposed. 1/ A bolder step would be to allow unsubsidized competitive ser- vices on those routes where the municipalized service is operating with an average waiting time in excess of, say, seven minutes.

1.79 These routes would not be the dense ones on which the large bus presumably has a relative advantage, but it is doubtful if the bus trade unions would sit idly by as their fiefdom is clipped.

1.80 Yet another possibility can be developed organizationally from the existing taxi service. Suppose that any taxi owner--and most are owner- drivers--was allowed to run a shared taxi service similar to that run in Belfast. It is then only a step to authorizing that vehicles larger than conventional taxis may be used, giving rise to the jitney type of operation.

1.81 In practice there have been many movements toward the emancipation of urban transport from the large municipalized bus undertaking. All have been much constrained by the thicket of regulations and the restrictive practices that such control has fostered. Groups of commuters or perhaps whole small communities have operated their own private bus commuting service with part-time or amateur drivers often at embarrassingly high profits and low fares. Similarly, firms are finding it in their interest to organize trans- port for workers through van pools and other administrative arrangements. Whether these measures, technologies and systems would be adopted if there were considerable freedom for transport to respond to the demands of passengers is not clear. The particular forms may be the product of the regulations.

1/ Bus experts and administrators often say--"but who would want to provide service on such unprofitable routes?" Clearly if these routes are inherently unprofitable, t4ere is no need to prevent entry! - 28 -

1.82 In many modes of transport,recent years have seen a pronounced shift in policy away from the controlledcartel or monopoly or regulated "competition"toward freedom to enter the industry,to charge competitive fares and to arrange operationsaccording to demand. Internationalair transport (and domestic in the United States) and trucking are two examples where competitionand freedom have brought remarkable benefits. It is at least likely that substantialbenefits would flow away from competitionin urban passenger transport.

II. THE BENEFITS OF MINIBUSES--THECASE OF KUALA LUMPUR

A. The Minibus Policy

2.01 The introductionof 400 minibuses in Kuala Lumpur (population approximatelyone million) enables an analysis of the net benefits of the policy changes. Minibuses, although accounting for only some 17% of total bus capacity, carried half the bus passengerkilometers. On conservativeassump- tions it is shown that the value of the benefits was US$10 million annually or almost one percent of income per capita. This value is surprisinglylarge and suggests that such a policy might have a similar payoff in other countries.

2.02 Debates about the desirabilityof allowing minibus operation have usually ended either inconclusivelyor with the general feeling that anything the minibus can do the big bus can do better. On many occasions the evidence which has been adduced to support these conclusionshas taken the form of calculationsby knowledgeableoperators of a bus system of the costs they would expect if they were required to operate a minibus system. The figures always show that the costs of a minibus are much higher than those of a big bus--both in terms of seat km and passengerkm. The presumption is that except for certain lightly trafficked routes, the minibus is not appropriate. (It is rather like the exercise which is common among engineeringstudents; they show that a bumble bee cannot conceivablyfly!) Furthermore,it is supposed that in general the value of the reduction in waiting time occasioned by the use of minibuses is likely for the vast majority of people to be less than the cost premium involved.

2.03 Impressive though these a priori argumentsmay be, they remain based on preconceptionsrather than reality. It would be instructiveto confront these propositionswith real evidence of the decisionsof individualsin the real world. Fortunately,we can now do just this. Minibuses were introduced in sizable numbers in Kuala Lumpur in 1976. Almost 400 were licensed to compete with the existing 700 stage buses which eventuallyreduced to 520-550. The rationale for this change in policy by the Malaysian government was three- fold: the need to expand bus transportcapacity; the belief that minibus services would reduce urban congestion by inducing existing and incipient motorists to forego their private car trips; and because it was planned to - 29 -

introduce an area road pricing scheme rather like that which had been in operation in Singapore since 1974 as it was perceived that additional public transport capacity would be required. 1/

2.04 The introductionof 400 minibuses was achieved in less than two years and during this period the price of the alternative forms of transport-- buses, taxis and (the costs of) private cars--remainedsubstantially the same. Bus fares were regulatedand taxi fares, nominally controlled,remained more or less the same. By 1976, most of the inflationhad been eliminated from the Malaysian economy and we can assume that there was no substantial change in the cost of private motoring, including the congestion costs in Kuala Lumpur and the cost of parking. The fare on the minibus was also regulated at M$0.40 cents irrespectiveof distance,whereas the bus fare is graduated according to the length of the trip, and similarlythe taxi fares depend on the length of the journey. 2/

2.05 The stage buses were given franchisesto operate certain routes and each operator can only operate on a specified route. In principle,minibuses were also required to operate on certain routes; in practice,however, they ply routes according to their judgment of the profitabilityof the traffic.

2.06 It is fairly clear that no passengerwill value a trip by minibus at a price which exceeds the price of a taxi. This gives us then an upper limit to the demand price for minibuses. Similarly the price of taxis, in the absence of minibuses, must give an upper limit to the demand price for bus trips. On the introductionof minibuses, however, it is clear that the demand curve for buses will fall, and in particular,the maximum demand price will fall to the region of the regulated fare on the minibus alternative.

2.07 These conditions enable us to characterizethe introductionof minibuses in Kuala Lumpur in the form of demand curves shown in Figures 2.la and 2.lb. Suppose that pt is the price of a taxi passenger trip (or alter- natively a private car passengertrip. The initial demand curve for buses Db (p >pt) is drawn on the assumptionthat there are no minibuses--hencethe presumption that the price of a minibus trip exceeds the price of a taxi journey. The minibus fare is regulated and fixed at p --and we assume that minibus trips are produced under conditions of constant cost equal to the fare of p . The regulatorsare "experts" and have managed to fix the fare so that the bus companies earn "normal" and not excess profits. The demand for minibus passenger trips D (P ) is determinedmainly by the price of the substitutebus passenger trip anQ is shown in Figure 2.lb. Agaiu the costs of a minibus passenger trip are assumed constant and equal to pm .

1/ The licensing of the minibuses and other policy measures were part of the Second Kuala Lumpur Urban Transport Project. This project was, in part, financed by a World Bank loan; however, no project funds were expended on the provision of minibuses. They were financed entirely by the private sector. 2/ As one would expect the average trip length of minibus passengers is 8.2 km (5.12 miles) compared with 3.9 km (2.44 miles) by stage bus. - 30 -

Figure 2.1 DEMAND FOR PASSENGER TRIPS

2.la: BUSES 2.lb: MINIBUSES

Pt Pt

p

Pb I I i I PaS

I I passenger passenger I: | trips I trips

2 1 ml (=O) H2

2.08 The number of bus passenger trips when minibuses are banned, is B1. When minibuses are allowed to operate at a fare of p_, which is at least as great as the marginal cost, they carry M2 passengers. The bus demand declines to Db(Pm) and the number of passengers trips to B2.

B. AlternateRegulatory Systems and Effects

2.09 In practice,however, in Kuala Lumpur, regulationsare imposed not only on the fares but also on the number of vehicles (each of which must be authorizedby an official license). The effectivenessof the regulatory legislation,therefore, depends on the extent to which, and the direction in which, thecantrolledprice and quantity deviate from what would have been the values in their absence. First consider the case of the stage buses. It is frequentlyargued by the operators that the regulated fares are below the marginal costs of operation. Although such allegationsmust be taken with a large pinch of salt, there is evidence that profits were very low in 1972-73 and in 1975-76 even with what are by normal standardshigh load factors.l/ A flat rate increase of M$0.05 (about US$0.02) on all fares in 1977 was approved to make services profitable and induce investmentin new buses. 2/ The author- ities were worried about the decline of bus services and the withdrawalof vehicles even at the new fare. This suggests that, if not actually below

1/ This emerges from an analysis of the publishedaccounts. 2/ After the event, the average fare was a little more than M$0.20 (about US$0.08) which implies a simple increase of about 33%. However, this is an inappropriatecomparison because the compositionof trips changed. - 31 - cost, the regulated fares were not appreciably above cost for the normal run of services. These arguments suggest, therefore, that the assumption that the regulated fares are near cost is appropriate. Furthermore, the quantity constraints imposed by the licensing authority do not seem to be binding; it is the regulated fare that constrains the supply. 1/

2.10 This has not been the case with minibuses since their introduction in 1975. The regulated fare of M$0.40 per trip was well above the marginal cost of providing that trip which is probably about M$0.21-0.22 (Annex 2). There has been an insistent pressure for the issue of more licenses. (In 1975 alone applications for licenses for 2,200 minibuses were received.) And this is clearly because substantial profits, far above normal, are to be earned by operating minibuses. A calculation of the level of profits, and an attempt to identify "supernormal" profits will be found in Annex 3.

2.11 If the minibus marginal costs exceed the regulated maximum bus fare and if this regulated fare applies to all bus journeys, including those by minibus, no minibus service could operate profitably. Thus there is no need for regulations excluding minibus operation; the fare control--exercised over minibuses as well as buses--is adequate to achieve that end. E

2.12 In Figure 2.lb there is no obvious account taken of the rationing constraint. But the licensing control is taken into account implicitly in the demand curve D (p). The demand curve D (p ) in Figure 2.2 represents the demand for minibuses when the quantity oT mKnibus licenses is constrained to a

Figure 2.2: DEMAND AND SUPPLY OF PASSENGER TRIPS BY MtlNIBUS UNDER DIFFERENT REGIMES OF REGULATION

P1

PM3

Pm2

passenger trips M2 M4 by minibus

1/ There is much variation between routes, but this statement is meant to apply to the generality of bus services. - 32 - particular figure (400). The demand curve D (p ) represents the equilibrium demand curve when there is free adjustment oA the quantity of minibuses and the fares to serve that demand. The former demand curve D lies below the equilibriumdemand curve D because the licensing constraint induces operators to crowd their buses and debase the quality of their service (for example, with numerous standing passengers). And, of course, queueing times for the minibuses would be far longer than under the equilibriumfree entry system. If the minibuses completelyadjusted their quality of service so that the demand price was equal to the regulated fare, the demand curve D (pb) would pass through the point given by the regulated fare and the observed number of passengers. Or, to put this point another way, the regulatedfare and the constraint on the number of minibuses induces the operators to crowd their vehicles and give longer waiting times than would be the case under free entry and fare conditions;and these lower-qualitycharacteristics are taken into account in the consumers'valuation of minibus passenger trips, as shown by the regulatedfare. In the evaluation of this paper, the value of introducingminibuses has netted out the disbenefit that arises from this continuationof both quantity (at 400 minibuses) and price controls (at M$0.40 per trip). 1/ It is necessary to.emphasize, however, that the heavier loadings on the minibuses, which such queueing and crowding entail, contributessub- stantially to the excess profits of minibus owners. Thus the net loss due to regulation is not as large as the apparent loss to passengers.

2.13 It is worthwhile reflecting on the evaluationof alternativeregimes. The first is obviously the case of complete freedom of entry and no regulation of fares. Then the relevant demand curve is D (p ). Thus yith the assumption that marginal costs are constant and equal to Fhe price (p ), this free entry and fare industry will result in H4 minibus passenger tripss.

2.14 The second case occurs if the fare controlwere removed but the li2ensed quantity restrictionsremain. Then the free fare would rise above p as the number of trips is reduced below M 2 Minibus operators would find it advantageousto increase the fare and reduce their passenger loads. But with the same number of minibuses the reduction in load factorswould give rise to an improvementin the quality of the service so the demand curve would rise above D (pb). With the restrictionon the quantity of minibuses, however, the quality of servicewould not rise to the level representedby the demand curve D (pb); the operatorswould find it advantageousto crowd their vehicles more than would occur under free entry conditions.

1/ In other words the rationingprocess causes a dissipationof the surplus due to the additionalqueues and crowding. See Yoram Barzel, "A Theory of Rationing by Waiting," Journal of Law and Economics, vol. 17, pp. 73-93, April 1974. - 33 -

2.15 The third case occurs if the fare control were retained at Pm but the quantitative restriction on vehicles were removed. Then the number of minibuses would expand and the load factors would fall until the marginal minibus breaks even at the new low load factors. And these would be lower than those implied by free entry and no fare control. Thus the quality of the servicewould be extraordinarily high with many empty seats and very rarely a full bus. The demand curve corresponding to this "high-quality/high-fare" service would lie above the demand curve D (pb) since the standard of service would be higher than would emerge in a free competitive situation.

2.16 As one would expect, the flat rate fare on minibuses in conjunction with the graduated fares on stage buses, gave rise to the longer passenger trips being taken by minibus. The average trip length by minibus was (8.2 km), more than twice that of the stage bus (3.9 km). Since the average fare on stage buses was M$0.21 and on the mninibuses about M$0.40, this implies that the fare per kilometer was roughly the same after the minibuses were intro- duced. 1/

2.17 In principle, one should draw figures such as 2.la and 2.lb for each length of trip since the price ratio between bus and minibus varies systemati- cally with distance. Where the ratio minifare/stage bus fare is lowest on long distance trips, one would expect the switch to minibus to be high. On the other hand, for short distances where the minibus premium is large, one would expect stage buses to retain much of their traffic. Due to lack of data, these refinements have not been proved in the ensuing calculations of bene- fits. 2/ The figures must be interpreted as containing errors due to this averaging procedure, but it is unlikely that such errors are sufficiently important to qualify seriously the conclusions. But the distribution bus trip length between minibus and stage bus has considerable implications for policy which will be raised in the last section this paper.

C. Estimates of Benefits

2.18 The next task is to fill in some of the numbers on the figures.2 The regulated price of a minibus trip, whatever the distance, was M$0.40 (=p ). 3/ Corresponding to the fare one observes that there are 400 minibuses and That the total number of trips per day is about 200,000. This provides then one point on the demand curve for minibuses.

2.19 An upper limit to the demand price for a minibus can be obtained by considering the price on the superior alternative, the taxi cab. The equiva- lent taxi fare for an average minibus trip would be on the order of M$0.80 to M$1.00 per person, assuming three persons (an effective maximum) share the

1/ It is difficult to be precise about such calculations because of conces- sionary fares and some split journeys by minibus. 2/ There are no detailed fare and distance distributions available. 3/ There were reports that some operators tended illegally to "split long trips" into two separate trips and collected M$0.80. However, such practices were thought to be quantitatively important. - 34 - cab. If the fare of a minibus trip were fixed at M$0.80, there would no doubt be still some business for the minibuses; but we have assumed that, at this fare, the demand falls to zero, giving an additional point on the demand curve. 1/

2.20 The simplest assumption about the shape of the demand curve is that it is linear and passes through the two points. This is the assumptionthat is used to evaluate the consumer'sevaluation of the minibus service. However, it is worth reflecting on alternativeshapes and examining the sort of results one would obtain. One of the commonest hypothesis about the shape of the demand curve is that it has constant elasticity,and we may attempt to draw a curve with such a constant elasticitythrough the point (M$0.40, 200,000 passenger trips a day). But what elasticity? As can be readily calculated, the postulated linear demand curve has an elasticity of (minus) unity at the observed point and the elasticityincreases as the fare increases.2/ An elasticity of unity is probably on the (absolute) low side and one may examine an alternativehypothesis that the curve has a constant elasticityof (minus) 3 until it reaches the upper limit of the fare of M$0.80; then it becomes horizontal. The constant elasticity curve is shown in Figure 2.3. It will be

Figure 2.3: DEMAND FOR MINIBUS TRIPS

0.80 --

0.60

constant elasticity 0.40 _--- - ^ *-3 0.30

passenger trips -___ _-_ ' >per day 100,000 200,000

1/ This assumption biases the case against the minibuses,but see next page. 2/ The Transport and Road Research Laboratory study, "A TheoreticalModel for Estimating the Effects of Fares, Traffic Restraint and Bus Priority in Central London," by R.H. Oldfield (TRRL Laboratory Report No. 749, 1977) used an elasticitywith respect to generalizedcost of unity (page 7) based on "the rather inadequateevidence available." But the study warned that "recent work" suggested that the elasticitymay be rather higher than one (presumablyin absolute terms). - 35 - noted that it cuts the linear demand curve at 100,000 trips per day, and finally meets the horizontal line through M$0.80 at a little more than 25,000 trips a day. It is clear that the area under the truncated constant elasticity demand curve is about the same as the area under the linear demand curve. The refinements of nonlinearization are probably not worthwhile--but any reader may choose his own assumptions and recalculate.

2.21 The consumer surplus can now be computed. It is clearly: 200,000 x (0.80 - 0.40)/2 in M$ on a daily basis, or M$40,000 per day. If this be grossed up on the assumption that there are only 300 effective days per annum, then the total consumer surplus becomes M$12 million per annum (approximately US$5 million).

2.22 To find the total benefit of the authorization of minibuses, the excess or supernormal profits of the operators must be added. This value (Annex 3) amounts to M$32,000 for each minibus, therefore for 400 minis the total is M$12.8 million a year. The grand total of benefits is therefore M$24.8 million a year. And we can take it that, approximately, the total benefits are equally divided between the customers and the operators.

2.23 The total benefits of M$25 million (approximately US$10 million) are large. For a total non-school population of about 600,000, this repre- sents a sizable annual dividend--about US$17 per head, near one percent of their average income.

2.24 These figures are so large that skepticism is readily evoked. Some correlated figures may therefore be useful. First, on recent surveys of minibus and bus travel for center bound journeys, the ridership is split 35% on the minibuses and 65% on the stage buses with the minibuses, because of their flexibility carrying many more passengers during the rush hours. Overall, it seems very unlikely that minibuses carry less than 30% of all trips. This should be compared with availability of about 520-550 buses and 400 minibuses--or, translated into officially rated passenger capacity (at 58 passengers per bus and 16 per minibus), 31,000 and 6,400 respectively. Thus the minibuses have about 17% of the passenger capacity but carry 30-35% of the passenger trips.

2.25 More important is the fact that the average length of the passenger trip by minibus (8.2 km) is more than double that by the stage bus (3.9 km). With a split between minibus and stage bus of 30-70, this implies that the minibus and stage bus carried approximately the same number of passenger km. This is for the minimum limit of the estimate of minibus passenger trips. If we take the 35-65 split, then the minibus carried 53% and the stage buses 47% of passenger km. By any criterion, this is a phenomenal performance: 15% of capacity carries more than 50% of the passenger kilometers. 1/

1/ Passenger km are not homogeneous; it takes more resources to transport two passengers for 4 km than one for 8. But still the measure is impressive. - 36 -

2.26 Second, it might be thought that some account should be taken of the lower ridershipon the stage buses and in the taxis. However, this is not a "cost" of the introductionof minibuses except insofar as there is evidence of marked economiesof scale to the industry in these activities. All empirical evidence suggests that there are no economies of scale, in the conventional sense, either in the bus or the taxi business. 1/

2.27 Furthermore,the increase in frequencieswhich has occurred with the mixed bus-minibusservice has undoubtedlyincreased the demand for public . The passenger, attractedby the higher frequencyof service, will take minibus or bus, whichever gets there first. 2/ Thus there is no reason to impute any long-termloss to bus companiesor taxis for the loss of traffic; and although the increasedfrequency of joint minibus and bus capacity on the routes might be credited to the introductionon minibuses, it has been omitted from the calculations. It might be adduced that there are some short-term losses as the buses and taxis adjust to their new situation. The assets of both are readily sold on the second-handmarket and there seems to be no reason to add such costs. This is particularlyso in the case of the buses, where there was obvious evidence of over-age buses and frequent breakdowns in 1975. In 1978 the bus companies were prepared to purchase new buses, partly for replacementpurposes.

2.28 Third, one may claim that calculationof profits in excess of normal of the minibus owners is excessive. There are no reported data in Kuala Lumpur which can substantiatewhether this is anywhere near the right figure. In Hong Kong however it is widely reported that the rate of return on a similar operation (the public light buses) is in excess of 100%. Furthermore, the very large demand for the limited allowance of minibus licenses testifies to their high profitability. It is doubtful, therefore,whether the allowance shown in the figure for super-normalprofits is too high.

2.29 Fourth, there is the problem of externalities. It might be held that the M$25 million is partly and perhaps wholly offset by external and indirect effects which have now been included in the calculation. The most obvious one is road congestion. It is not clear which way one could expect congestion to go. On the one hand the provision of more frequent minibus

1/ In Kuala Lumpur all evidence suggests that the largest bus company, Sri Jaya, inherited from the British, is somewhat less efficient than other smaller companies. But I would not wish to rely solely on such evidence. See also "Economiesof Scale in Bus Transport,"N. Lee and I. Steadman,Journal of Transport Economics and Policy, 4, No. 1, 1970. 2/ On some routes where minibuses and buses operate, there has been some reductionin stage bus frequencies. Consequently,the passenger who has a low value of time may have to wait longer for a stage bus. However, he is more likely not to be bypassed by full'buses. But since these passen- gers apparentlywait for a stage bus and decline the offer of the mini- buses, one may presume that their value of time is very low indeed. - 37 - service would be expected to entice the erstwhile or incipient motorist, thus reducing congestion. And since they carried about as many passenger miles as the large bus, with a rated congestion effect of only half the bus per mile, it seems that the net effect on congestion may be expected to be favorable. Unfortunately, it is impossible to collect information which would confirm or deny this proposition. While it is true that congestion seems superficially to be certainly no worse than during the earlier days, there have been many other changes during this period (emergence from a slump, considerable imped- ing road works, an expansion of the private car fleet, etc.) which have prevented attribution of the effects to causes.

2.30 Fifth, there are externalities such as the "creation of jobs"-- whatever that much overworked term may be taken to mean. Compared with the standard bus, the minibus is a labor-intensive form of transit and substitutes labor for capital. (Often the minibus operators find it efficient to carry a fare collector who, one suspects, is usually a member of the family.) Further- more, the maintenance services of minibuses require less capital and can be performed by "shade tree" mechanics. Thus it seems likely that the minibuses provide more "productive" jobs in the private unregulatd sector than the simple stage bus system. It is also important to note that the employment is in the non-unionized sector, as distinct from the stage bus industry which is largely unionized, and furthermore, the minibuses provide an outlet for entre- preneurship on a small scale. The enterprising man who has a small amount of capital can enter the minibus business. Capital and ownership are spread throughout the population (and in particular the Malay population) rather than being concentrated in few hands. It would be interesting to know the income- employment characteristics of those who use minibuses, but unfortunately, no information is available.

2.31 It might be noted that it is difficult to disentangle the various elements that have contributed to this surprisingly large benefit of over US$10 million annually. One would suspect that much of it is due to the owner-driver form of incentives that induces the minibus operator to seek diligently where demand is high and to prune his costs to the unavoidable ones. 1/ Some part of the value is probably due to the technology of the smaller bus. But all this is conjecture. One may also speculate on what would be the result of allowing free entry into the minibus business. It is likely that minibuses would take over the majority of trips in Kuala Lumpur. And this would be in spite of the fact that Kuala Lumpur has, by any normal international standards, an efficient and cheap system of stage buses.

1/ In the Bank appraisal the minibuses were expected to carry approximately half the number of passengers they actually 'achieved. This forecast was done on the best assumptions of transport experts. But like the bumble bee ... - 38 -

D. Conclusionson Policy

2.32 The conclusion is that the introductionof minibuses produced an extraordinarilylarge benefit for both operators and passengers--abenefit which they shared more or less equally. The annual benefit is the equivalent of about a one percent increase in income for every adult in Kuala Lumpur, or about US$18 per capita. The constrainton the number of minibus licenses combined with the relativelyhigh regulated fare gave rise to very high load factors and substantialprofits for operators. With free entry, the fare would probably fall more than 40X, and it is likely that the minibus would be the predominantform of urban transport.

2.33 As could be expected with the flat rate regulated fare on minibuses, they have capturedmany of the long distance trips and left the stage bus with the shorter trips. This is the opposite of the normal role assigned to minibuses. One would conjecture that the greater frequencyof minibuses could be relativelymost advantageouson short trips. In other cities, such as Manila (Philippines),the minibuses ("jeepneys")concentrate at the short end of the urban travel market. But there both buses and minibuses have the same flat rate fare structure. So it appears that the more appropriaterole of minibuses in carrying short distance passengers is inhibited,or indeed largely precluded by the fare structureimposed by the regulators. They have been induced into the long hauls where the relative advantage is less. It is some sort of testimony to the virility and adaptabilityof the mini business that the minibuses have, nevertheless,much higher net benefits.

2.34 An obvious subject for further speculationis the deregulation of both fares and quantity restrictionsin the urban transportbusiness. The outline of the results of such a policy change is easy to see, but it is beyond the scope of this paper.

2.35 Kuala Lumpur was in no way a sitting duck for the minibuses. The stage buses were relativelyfrequent, ubiquitous and cheap, yet marginally profitable. Similarly, the taxi servicewas not tightly constrainedand fares were low. This suggests that the benefits were not due to the special circumstancespertaining in Kuala Lumpur. One may thus expect that similar acts of deregulationin other cities may produce benefits of a similar or even greater magnitude. 1/ One of the unique features of Kuala Lumpur is that the introductionof minibuses was achieved quickly in one large operation. Such a process is rare; governmentsusually like to proceed cautiously. And this often sets up sufficientresistance to stifle further deregulation. Herein is a lesson for reformers.

1/ A fuller discussion of the minibus operations in, for example, Manila, Belfast, Bangkok and Hong Kong is to be found in pars. 1.58-1.66. - 39 - ANNEX1

COSTS AND SIZES OF BUSES

1. It may appear to be obvious that the capital cost per seat of a large bus is less than that of a small vehicle since one is "spreading,"for example, engine and transmissioncosts over a larger carrying capacity. But this is not the case. A casual inspectionof prices will show that this is not so. Compare two extremes: The American 5-passenger-plus-driver auto- mobile and the standard AM 51-seat-plus-26-standeesbus. The automobile,air conditionedand appropriatelyfitted, would cost about $4,200 for a fleet purchaser. From recent contracts it appears that the bus, appropriately fitted, would cost nearly $100,000. Thus the cost per seat of the automobile is $840 compared with $1,960 for the bus. Even if one adds the places, the cost per passenger place is $1,300. (One should also take into account the differencesin length of life and in maintenancecosts before comparing the annual capital costs.)

2. The reason for the relative cheapness of the small vehicle is that they are produced in great volume, and most of the great productivitygains over the last half century have been due to mass production and line tech- niques. The bus, on the other hand, is produced by a low-volumebatch process. The above comparison understatesthe difference in capital costs per passenger place, since the standard automobiledoes not attempt to maximize its carrying capacity. If one adapted a small van carrying 10-12 passengers,the fleet cost would be on the order of $8,000--whichwould give an even lower cost per seat of $666.

3. These comparisonscover extremes. The more normal range of bus sizes 1/ was examined for 24 fleet purchases for cities in the United States between 1965 and 1970. 2/ This showed that the formula for the price of a bus was given by:

$ Price of bus = $29,000 + $1,000 (S-45) with S seats

(for price of buses from $24,000 to $33,000) where S is the number of seats

1/ Note, however, that we have not examined the articulatedbus such as the NEOPLANarticulated . In March 1975 this was quoted at $124,378 for 77 seats, 43 standing. It will be seen to be consistent with the American model.

2/ Department of Transportation "Economic Characteristics of the Urban Public Transportation Industry," prepared by the Institute of Defense Analysis, February 1978, pp. 4-74 et seq. - 40 - ANNEX I

4. Although the authors do not mention it, this equation can be true only over a certain range (e.g., certainly excluding S<161). The $1,000 a seat for capacity above 45 seats is indicativeof diseconomiesof scale--but it probably is numericallyan exaggerationsince the authors note that the larger the bus the more likely it has optional extras such as air conditioning and power steering. From a survey of the price lists, as distinct from fleet prices, it seems that an additional seat tends to add on the average about $800--but that is a rough figure.

5. Probably the only certain inference we can draw is that there is no evidence of scale economies of capital costs in large buses. The evidence indeed suggests that- there are pronounced diseconomies after a rather small size. - 41 - ANNEX 2

COSTS OF BUS AND MINIBUS

1. It is extraordinarilydifficult to assemble data on costs of both buses and minibuses on a comparable basis; the paucity of data is particularly acute in the case of the minibuses. Much published material on minibus costs is not the result of observationof going concerns. Some is "synthesized" (fabricatedwould be a more appropriateverb from supposed operating charac- teristics and unit costs. In many of the exercises the object seems to be to show that minibuses are uneconomic compared with the stage bus--and could not conceivably be profitable. As Amos pointed out, however, it is rather like engineers proving that a bumble bee cannot fly. And so running minibus operations according to the rules and conventions of the large stage bus monopolies would certainly lead to financial ruin. In practice the mini- operationsare different from stage-buspractices--as can be observed in the jeepneys in Manila (Table 1), the public light bus in Hong Kong (Table 2), and the minibuses in Kuala Lumpur (Annex 3),--even though they cater for a similar demand.

2. In many examples,among them Hong Kong, Istanbul,Kuala Lumpur and Manila, the size and type of minibus is mandated by the authorities(pars. 1.58-1.63). The vehicle must conform to regulatoryspecifications (par. 1.68). There is every reason to suppose that a single specificationof vehicle, especially if it is produced by a regulatoryauthority, is unlikely to be best for all services. For example, it is highly likely that there is a large latent demand for the services of a much lighter bus (such as the VW van with 10-12 seats) with lower comfortbut lower fares in both Hong Kong and Kuala Lumpur. The minibus is not the result of the free choice of the market, except in the case of Honduras (Tegucigalpa)(par.1.55). Similarlyin Istanbul (par. 1.66), Kuala Lumpur (par. 1.64) and in Manila fares are controlled which gives rise to a particular form of operation. In sum, one can never observe like against like operationsand consequentlythere will always be room for interpretation.

3. The costs of minibus operations,therefore, are likely to have an upward bias because: (a) the standard comparison is with the typical stage bus and its operationalcharacteristics; and (b) the lack of data are more pronoun- ced in the case of minibuses so that the analyst fills in with conservative assumptions. An illustrationof this is the propositionthat public light buses have a life of 5 years compared with the stage buses that have a life of 12 years. In the study no one could observe the life of minibuses and it is clear that 5 years is a short period. (Evidence on this appears in Turkey where, in Ankara and Istanbul, one may observe minibuses--or dolmus--which are American cars of the 1940-50s. Transport experts argue that, especially in view of the paucityof spare parts, such long lives are clearly uneconomic; but I am not persuaded that so many vehicle owners do not know their own interests better than the experts.) - 42 - ANNEX2

Manila

4. For only one city, Manila, have we data that are closely comparable for both bus and minibus--that is the jeepney. Table I compares the cost per mile and the cost per seat mile of the standard stage bus and the standard jeepney. The sources of these data are mixed; to some extent they are from survey material (this applies to the mileage) of a fairly reliable nature, and to some extent it relies on the best estimates of experts (this applies to tires, fuel consumption,lubricants and maintenance).

5. The first striking feature is that the annual depreciation and interest cost of the seat-mile of the jeepney is only about 40% of the compar- able figure for the bus seat-mile. This emphasizes the low capitalizationof the mini-operation. These low capital costs of the jeepney are to some extent offset by the higher operatingcosts--particularly in fuel and lubricants. It is indeed astonishingto find that the costs per mile of the lightweight Jeepney are only about 20Z less than the heavy bus; it seems that fuel costs are about 62% of non-wage operating costs of jeepneys. However, substantial taxes were imposed (10% on diesel and 38% on gasoline);when these are extracted,the fuel cost of the jeepney is about one-half that of the bus (2.784 cents a mile compared with 4.950). 1/

6. It will be seen that, in the totals, the cost, inclusiveof taxes but excludingwages-of the jeepney per seat mile is almost the same as the cost per seat mile of the bus.--The difference between them of .045 cents, or about 9%, is well within the errors of calculationof such figures. If, however, the fuel tax is extracted, then the cost per seat mile of the jeepney falls--perhaps significantly-below the cost of the bus--.4 cents compared with .465 cents, a difference of about 16%. 2/ Finally, consistentwith the remarks made in par. 3 above, it will be noted that this comparison is loaded against the jeepney in many respects. We list only two: (a) it is likely that there is less dead head running of jeepneys than buses and that the effective mileage of the jeepneys is greater than the figures given; (b) it is less than those of the large buses (they have been assumed equal in the comparison).

1/ There are still good reasons for supposing that the physical fuel consumptionof jeepneys has been overstated in these figures. 2/ In principleone should also subtract the vehicle duties to get the depreciationand other capital-relatedcosts free from tax. However, the adjustmentsare trivial compared to the fuel duty and hardly affect the comparison. - 43 - ANNEX 2

Table 1: MANILABUS AND JEEPNEY COSTS (1976 US cents)

Cost/Mile Cost/SeatMile Cost Item Bus Jeepney Bus Jeepney

Depreciation 7.4/a 0.75 0.128 0.054/a Interest 4 9 0.55 0.084 0.039 Maintenance 6.7/- 0.85 0.116 0.061 Tires 2.16 0.476 0.037 0.034 Fuel 5.5 4.48 0.095 0.32 Oil 0.384 0.17 0.007 0.012 Wages 9:6//-d 3.4 0.166 0.243 Management 0.44 - 0.008 - Total 37.08 10.68 .641 0.763 Total excluding wages 27.48 7.27 0.475 0.519

Total OperatingCosts 24.34 9.376 0.420 0.67 (fuel tax) (0.55) (1.70) (0.01) (0.12) Adjustments: (Factor cost of fuel) (4.95) (2.78) (0.086) (0.20) Total ExcludingWages (after adjustment to exclude fuel tax) 26.93 5.58 0.465 0.40

Note: Bus: 58 seats; Jeepney: 14 seats.

/a Depreciationestimates are based on a capital cost of US$30,666for a 55-seat stage bus with an expected 10-year life. The jeepney costs US$2,972 for a 14-seatvehicle lasting on average 7.5 years. Buses average 46,500 miles/year;jeepneys 50,000 miles. It is worth noting that the capital cost per seat mile of a bus is about 2 1/2 times that of a jeepney. 75% of depreciationcost is treated as dependenton the distance traveled,with the remaining25% determined by time in use.

/b Maintenancecosts for the two vehicle types are proportionate: 300 hours of labor time, and parts cost estimated at 10% of vehicle cost.

/c Wage costs of jeepneys and bus differ in wage rates paid and size of crew employed. A bus operates with a driver (@ P 5/hr.) and conductor (@P 4/hr.) Jeepneys employ only a drive @ P 2.5/hr.

/d License and insurancecosts are not provided. - 44 - ANNEX 2

7. It will also be noted that the labor costs of the jeepney are about one-third of those of the bus. This is partly because the jeepneys do not employ a separate fare collector and partly because bus drivers are LiEoughtto be more highly skilled (and of course they are unionized).

3. The fares are controlledat the same values for both jeepneysand buses. ie., P 0.25 for first km or US$0.033 for first 3.1 miles; and P 0.05 for each additionalkm or US$0.0096each additionalmile. At these fares, the buses appear to have considerabledifficulty in meeting their costs, but the jeepneysmanage to earn what many observers regard as adequate or sub- stantial profits. One of the explanationsis thought to be that the average Jeepney Journeyis considerablyshorter than the average bus journey. Al- though there is some doubt about these statisticson journey length, if true it would be consistentwith the jeepneys charging more for their superior service.

Calculatingm ( R ) = a + b x for Manila:

The wage costs per mile of a bus and jeepney are respectively9.6 and 3.4 US cents, and these correspond to 58- and 14-seatvehicles respectively. Fitting a linear relationshipto these points gives a slope of 0.141 cents a seat, and an interceptof 1.42 cents. Thus if x is the number of seats per vehicle and y is the wage cost (in U.S. cents) per mile per vehicle

y = 1.42 + 0.141 x

Divide x by 58 to giveA , and suitably standardizingthe intercept,one obtains

m = 0.148 + 0.852g

Hong Kong

9. There are fairly reliable data on the public light buses (PLBs) but only for 1972. The first striking feature in comparisonwith the Manila figures is the small value for depreciationper seat mile (including interestcharges) compared with the jeepney, i.e., US$0.16 in Hong Kong and M$0.092 in Manila. This reflects the sturdier nature of the PLB and the fact that it was designed as a minibus; it is also misleading to calculatethe comparison per seat since the PLB has much more potentialfor standing passengers than the jeepney. (One interestingcomparison is that the cost of fuel per seat mile for the PLB is only about 60% of the cost of fuel for the jeepney.)

10. The costs, excludingwages per seat mile are 0.711 US cents in Hong Kong compared with 0.519 for the jeepneys in Manila. Whether this difference in costs measures the difference in comfort and standing capacity is a matter which can be pursued no further here. It is clear, however, that the two costs are broadly comparableor at least not dramaticallyinconsistent. - 45 - ANNEX 2

Table 2: HONG KONG PUBLIC LIGHT BUSES (PLBs) COSTS (1972 US$ plus 20%)

/a Cost Item Cost/Mile Cost/Seat Mile

Depreciation 2.30 0.164 Maintenance 2.56 0.183 Tires 0.30 0.021 Fuel 2.78 0.199 Oil 0.08 0.006 Wages 7.10 0.507 License 1.25 0.089 Insurance 0.68 0.049 Management - - Total 17.05 1.218 Total ExcludingWages 9.95 0.711 Total Operating Costs 12.82 0.916

/a 14-seat bus.

Other Cities-SanJose, Kuala Lumpur

11. In San Jose (Costa Rica) and in Kuala Lumpur there are fairly good data on stage bus operation, but none on minibus costs. These are broadly consistent with the Manila figures; thus in 1974, the cost per seat mile of the 50-seat bus in San Jose was 0.557 US cents, and in Kuala Lumpur the 44-seat bus cost 0.69 US cents in 1976. Allowing for inflation, these values are virtually the same and are within 10% of the values for Manila.

Table 3: KUALA LUMPUR: STAGE BUS COSTS (1978 US$)

/a Cost Item Cost/Mile Cost/SeatMile

Depreciation 4.17 0.094 Interest 1.7 0.027 Maintenance 5.4 0.123 Tires 5.4 0.123 Fuel 3.4 0.078 Oil 0.56 0.081 Wages 7.4 0.168 Insurance 1.08 0.025 Management 1.35 0.031 Total 30.46 0.69 Total Excluding Wages 23.06 0.525 Total OperatingCosts 23.86 0.543

/a 44-seat bus. - 46 - ANNEX3

PROFITS OF MINIBUSES

1. The expected accounts of a minibus operator were calculated for the appraisal of the Second Kuala Lumpur Urban Transport Project. 1/ These data assure a one-shift operation and the following operating data 2/ at 1975 speeds:

Item

1. Days of operation per year 300 2. Hours of operationper day (effective) 8 3. Average speed, mph 15 4. Average occupancy,passengers 7 5. Average passenger trip length, miles 4 Number of passenger trips carried: 6. per day 210 7. per year 63,000

Revenues 8. Daily revenues (M$) 84 9. Annual revenues (M$) 25,200

Expenditures(annual)

Fixed 10. Capital cost (purchase price M$32,000), 5-yrs @ 12% 8,900 11. Vehicle insurance 1,400 12. Licenses and road tax 1,200 13. Drive (M$7/day plus 80% overhead for Employee Provident Fund, etc.) 3,800 14. Total fixed expenditure 15,300

Variable 15. Fuel (diesel; 33 mpg @ M$1 per gallon 1,100 16. Maintenanceand repairs (M$25 per 1,000 miles) 900

17. Total variable expenditure 2,000 18. Total expenditure 17,300

Profit 19. Items 9-18 7,900

1/ Second Kuala Lumpur Urban Transport Project: Appraisal Report World Bank, Washington, D.C. 1976 (processed) 2/ See also pars. 1.58-1.66. - 47 - ANNEX 3

2. We know that the number of passenger trips per day is approximately 500 (420 for 365-day year) comparedwith the estimate of 210. This increased number of trips was achieved largely by increasing the average load of passen- gers from the assumed 7 to a figure probably nearer to 14 for the one (8 hour) shift. 1/

3. If we suppose that the only change from the data shown in par. 1 is the doubling of the average load from 7 to 14 passengers,the approximate result would be a doubling of annual revenue--fromM$25,200 to M$50,400. Subtractingtotal expenditureof M$17,300 from this, we obtain a profit of M$33,100.

4. This represents a more than 100% rate of profit on the investment of M$32,000. And clearly this is a lower bond because we have not yet taken into account the additional80 passenger (500-420)trips a day which, on the average, minibuses achieve--whichwould bring in a yearly revenue of M$9,600. If variable costs increasedproportionately, i.e., (80/420)x 2,000, this would give an additional cost of M$381--givingan additionalnet revenue of M$9,219. Thus the best estimate of total annual profit is M$33,100 + M$9,219 or approximatelyM$42,000. This represents a rate of return on capital of over 130%.

5. Clearly these rates of profit in a business such as minibus operation far exceed "normal" values. "Normal" profits in regulatedurban passenger transportare often calculatedat 15-18% of revenues--whichwould give about 12-14% return on investment. The appraisal report concluded that the return would be about 31% of revenues and 25% on investment--andthese rates were clearly regarded as greater than normal.

6. From all this discussion it appears that excess profits (over and above normal profits) are at least 100% of capital, or approximatelyM$32,000 per annum, which implies a normal rate of profit of about 30% on capital. This figure was used in the evaluation.

7. It will be noted that this figure is broadly consistentwith the rate of profit on one other similar minibus operation--thepublic light buses in Hong Kong. There it is reported that profit rates are "more than 100%." 2/ This suggests that, if there are approximatelyconstant costs in the minibus industry,then free entry and no fare regulationwould result in the competi- tive fare of about US$0.21 or US$0.22.

1/ During the three-hourperiod 0630-0930, 17.1 passengerswere carried and for the peak hour over 18 passengers. Thus it is assumed, as the data show, that minibuses operate intensivelyduring the peak, the 8-hour shift will cover both peaks with approximately17 carried during the two 3-hour peaks and 10 during the two off-peak hours. 2/ Barden, S.A. and Seneviratne,J.R.V., Public Light Bus Operation Survey, 1972, Public Works Department,Hong Kong, June 1973. - 48 -

REFERENCES

Amos, B.: "Shared Taxis in Belfast: An Enquiry into the Relevance for ParatransitDevelopment." Presented at PTRC Summer Annual Meeting, 1978. (PTRC Education and Research Services,Ltd., 109 Bedford Chambers,King Street,London, W.C. 2.)

Barden, S.A., and Seneviratne,J.R.V.: Public Light Bus Operation Survey, 1972, TechnicalReport No. 115. Traffic and Transport Survey Division, Public Works Department,Hong Kong, June 1973.

Barzel, Y.: "A Theory of Rationing by Waiting." Journal of Law and Economics, vol. 17, April 1974, pp. 73-93.

Bertrand, T. J.: "CongestionCosts in a TransportationSystem with an Application to Bangkok." Journal of Transport Economicsand Policy, XII (3), September 1978, pp. 244-271.

Britton, F.E.K.: AlternativeTransport Technologiesfor Developing Countries, OECD, DevelopmentCentre, CD/TI (74) June 1974.

Paratransit: An InternationalSurvey of Innovative Bus, Taxi and AutomotiveTransit Arrangements(550 pp.) EcoPlan, Paris, 1977.

"ParatransitDevelopments in the OECD Region." (Extractssubmitted to Workshop on ,Paris, 1977.)

Butler, R.: Bus Operation in the Colony of Hong Kong: A Study of Public and Private Enterprise. (processed,1970.)

Coase, R.H.: "The Problem of Social Cost." Journal of Law and Economics, vol. 3, October 1960, pp. 1-44.

Fouracre, P.R.: "IntermediatePublic Transport in Developing Countries." Transport and Road Research Laboratory Report No. 772, TRRL, Crowthorne, Berkshire, 1977.

Fouracre, P.R., and Maunder, D.A.C.: "Public Transport in Chiengmai, Thailand." Transport and Road Research Laboratory SupplementaryReport No. 285, TRRL, Crowthorne,Berkshire, 1977.

"Urban Public Transport in Chiengmai, Thailand." (Reprint) Traffic Engineeringand Control, vol. 18, No. 5, May 1977.

Freeman, Fox, and Associates.: Metroplan: MetromanilaTransport, Land Use and DevelopmentPlanning Project, Final Report, London, July 1977.

Grava, S.: "The Jeepneys of Manila." TraffictQuarterly, vol. 26, No. 4, October 1972.

Hong Kong Standard,June 1, 1978. - 49 -

Jacobs, G.D., and Fouracre,P.R.: "IntermediateForms of Urban Public Transport in DevelopingCountries." (Reprint)Traffic Engineering and Control, vol. 17, No. 3, March 1976.

"Urban Transport in DevelopmentCountries." (Reprint)Traffic Engineeringand Control, vol. 15, No. 20, December 1974.

Kirby, R.F.: Paratransit,Washington, D.C., The Urban Institute, (Ur 91000), 1974.

Lee, N., and Steadman, I. "Economiesof Scale in Bus Transport." Journal of TransportEconomics and Policy, vol. 4, No. 1, 1970.

Martin, B. and Warden, C.: TransportationPlanning in Developing Countries, The Brookings Institution,Washington, D.C., 1966.

Mohring, H.: "Optimizationand Scale Economies in Bus Transportation." American Economic Review LXII (4), September1972, pp. 591-604.

TransportationEconomics, Ballinger, Cambridge,Mass. 1976.

Nash, C.A.: "ManagementObjectives in Bus Transport." Journal of Transport Economicsand Policy, XII (1), January 1978, pp. 70-85.

Oldfield,R.H.: "A TheoreticalModel for Estimating the Effects of Fares, Traffic Restraint and Bus Priority in Central London." Transport and Road Research LaboratoryReport No. 749, TRRL, Crowthorne,Berkshire, 1977.

Sanli, H.I.: Dolmus-MinibusSystem in Istanbul. A Case Study in Low Cost Public Transport,Istanbul. University of Istanbul, 1977 (processed).

Smeed, R.J., and Wardrop, J.G.: "An Exploratory Comparisonof the Advantages of Cars and Buses for Travel in Urban Areas." Instituteof Transport Journal, vol. 30, No. 9, March 1964.

U.S., Departmentof Transportation,Urban Mass Transit Authority: Characteristicsof Urban TransportationSystems. Washington,D.C., 1974.

National Academy of Sciences: Noise Abatement: Policy Alternatives for Transportation,prepared by the National Research Council, Washington,D.C., 1977.

Voorhees, A.M., and Associates, Inc.: San Jose MetropolitanArea TransportationStudy, Final Technical Report, Republic of Costa Rica, Ministry of Public Works and Transport, McLean, Va., March 1976.

Wells, J.D., et al.: Economic Characteristicsof the Urban Public TransportationIndustry. U.S. Department of Transportation,prepared by the Instituteof Defense Analysis,Washington, D.C., February 1978, pp. 4-74.

Wilbur Smith and Associates.: Hong Kong ComprehensiveTransport Study, Washington,D.C., 1976. RECENT PAPERS IN THIS SERIES

No. TITLE OF PAPER AUTHOR

295 The Technology of Rural Development J.P. Mclnerney (consultant)

296 The Financial Cost of Agricultural Credit: C.D. Datey (consultant) A Case Study of Indian Experience

297 Agricultural Sector Planning Models: A.C. Egbert A Selected Summary and Critique

298 Textbooks and Achievement: What We Know S.P. Heyneman J. Farrell, A. Sepulveda-Stuardo (consultants)

299 An Economic and Social Analysis of the Chao C. Bruce Phya Irrigation Improvement Project II Y. Kimaro

300 Two-Studies of Development in Sub-Saharan S. Acharya Africa B. Johnston (consultant)

301 The IntermediateSector, Unemployment, and W.F. Steel The Employment-OutputConflict: A Multi- Y. Takagi (consultant) Sector Model

302 The Economic Theory of the Household and D. Chernichovsky Impact Measurement of Nutrition and Related Health Programs

303 Trade Restrictionsand InternationalPrice M. Bale Instability E. Lutz

304 IntergovernmentalFiscal Relations in R. Bird (consultant) Developing Countries

305 A Programming Approach to Fertilizer A. Choksi Sector Planning A. Meeraus

306 The Foreign Exchange Gap, Growth and K. Dervis IndustrialStrategy in Turkey: 1973-1983 S. Robinson

307 The Importanceof Risk in Agricultural P. Hazell, R. Norton Planning Models M. Parthasarathy. C. Pomareda (consultant)

308 Guidelines for School Location Planning W. Gould (consultant)

309 Growth and Poverty in Developing Countries M. Ahluwalia, N. Carter H. Chenery -2-

No. TITLE OF PAPER AUTHOR

310 Teacher Training and Student Achievement T. Husen, L. Saha in Less Developed Countries R. Noonan (consultants)

311 Optimum Economic Power Supply Reliability M. Munsainghe M. Gellerson (consultant)

312 Intra-IndustryTrade and the Integration of B. Balassa Developing Countries in the World Economy

313 Export Promotion Policies B. de Vries

314 The Changing Composition of Developing H. Chenery Country Exports D. Keesing

315 Urban Growth and Economic Development in the M. Cohen Sahel: Prospects and Priorities

316 World Trade and Output of Manufactures: D. Keesing Structural Trends and Developing Countries' Exports

317 Cuba: Economic Change and Education Reform M. Carnoy, J. Wertheim 1955-1974 (consultants)

318 Sources of Fertility Decline: Factor Analysis R. Faruqee of Inter-CountryData

319 Educational and Economic Effects of Promotion W.D. Haddad and Repetition Practices

320 Small Farmers and the Landless in South Asia I.J. Singh

321 Fruit-and Vegetable Exports from the R.D. Hunt MediterraneanArea to the EEC

322 Ability in Pre-Schoolers,Earnings, and R. Grawe Home-Environment

323 Priorities in Education: Pre-School; M. Smilansky (consultant) Evidence and Conclusions

324 Tropical Root Crops and Rural Development T.J. Goering