The Transport Research Laboratory is an executive agency of the Department of Transport. It provides technical help and advice based on research to enable th~ Government to set standards for highway and vehicle design, to formulate policies on road safety, transpu:t and Ithe environment, and to encourage good traffic engineering practice. TRL also sells its services, acting as contractor, consultant or providing facilities and staff on a fee-paying basis for customers in the private sector. TRUS expertise ranges from the construction of highways, bridges and tunnels, to Y all forms of road safety, traffic controi and driver behaviour. For instance, highways and structures research at TRL develops improved materials and methods which enable earthworks, roads and bridges to bl~ designed, built and maintained more cost-effectively. New ways of ~einforcjng earth can cut construction costs, while bridges can l)e modified to re(luce corro~ion. Road surfaces developed at TRL can reduce nc)ise and cut wel-weaihcr road spray from lorries by 90 per cent. Safety research varies from monitoring the incidence of drinking and driving and devising ways of reducing it, to improving junction designs and cooperating with European partners on new standards for improved impact protection in vehicles. Traffic research seeks to make the most of existing roads by, for instance, improving traffic signal coordination and devising systems which help drivers avoid congestion. Other research looks at the effectiveness of parking controls and improved crossings for pedestrians. TRL research also informs Government transport policy by studying, for example, the effects of bus deregtiation and how land use interacts with the road and rail transport system. TRL employs around 600 scientists, engineers and support staff. Its headquarters are at Crowthorne, Berkshire where its facilities include a 3.8km test track, a structures hall where bridge structures can be stressed to breaking point, a facility for carrying out accelerated tests on road structures, and advanced computer systems which are used to develop traffic control programs. TRL Scotland is ~ situated in Livingston, near Edinburgh, where the staff are concerned mainly with research and advisory work in the fields of ground engineering, bridges and road pavements. This unit has responsibility for all TRL work in Scotland. ~ A large proportion of the research is sub-contracted to industry, consultants and universities. The Labori~tory also collaborates with local authorities and other organisations within Europe and elsewhere. In addition, TRL expertise is I provided to developing countries as part of Britain’s overseas aid programme. I For more ! information: TRL Public Relations, 0344 ~0587 I I Old Woklngham Road CroMhorne Berkshire RG11 6AU PROJECT REPORT S4 OverseasDevelopmentAdministration 94 Mctoria Street London SwlE5JL PAKISTAN ROAD FREIGHT INDUSTRY: AN ANALYSIS OF TARIFFS, REVENUES AND COSTS By J L Hine and A S Cl~iIver rSubsector. Transport ‘~ Subject: National Transport Plannl;lg Theme: Increased Efficiency of Nai!snal Transport Systems ,, Project title: Transport Efficiency/Freighi Project reference: ODA R5593 I -~ .-. This document is an output from an ODA-funded research project, carried out for the benefit of developing countries. Crown copyright 1994, The views expressed in this report are not necessarily those of the Overseas Development Administration or the Department of Transport. Overseas Centre, TRL, 1994 ISSN 0968-4093 ) ‘1 ( CONT~NTS~~ ,, ., ,., , ,:- =acutive Summa~ EXECUTIVE SUMMARY . . . . . II The research described In this report is part of a major study cetied out by the Overseas Unit of the Trens~rf Research Laboratory in cooperation with the National Transport Reseamh Centre of Pakistan in Islamabad. Fmm 1985 to 1987 a wide range of data was collected on Pakistan’s private mad freight transpoti industry includng information on freight tariffs, vehicle utihsatlon, operating costs, vehicle ownership and management, and road roughness. A general description of the industry has been presented in Research Repoti 314 and an analysis of vehicle time utifisatlon in Research Report 333. This report describes an analysis of tariffs, revenues and costs. LrrTariff = 1.74 + 0.24 Direction Variable (to Karechi=O, fmm Karechi=l) + 0.447 Ln Distance (km) + 0,198 Ln Time (hre) + 0.282 Ln Load Weight (tonnes) + 0,317 Ln New Vehicle Value (1000s Rs) 1. 2. 3. 4. Introduction Date sources 334444 Tariff trends and seasonsfity 3.1 The date 3.2 General tariff trends 3.3 Tariff trends by direction .. To examine the effect of large differences in mad rough. ness on revenues a comparison was made between a very rough route In the Mekran area of Baluchlstan (roughness 13,000 m~m on the El scale) and routes on the rest of the main road network (roughness 4,800 mtim on BI scale). On the rough route, revenues per vehicle kllometre were estimated to be batween 10 per cent and 24 percent higher depending on vehicle type. However because of the lower loads carried on the rough route, revenues per tonne km were between 63 per cent and 79 per cent higher cn this mute mmpared with the main road network. There is a need to improve the understanding and modelfing of the factors which determine freight costs and tariffs for a wide range of planning purposes. ,41thougha A cmss-stional analysis of tariffs and other data 6 4.1 The date 6 4.2 me explanation of tariffs 6 number of major stuties of vehicle operating costs have been carried out it has been found that the ;ost relationships can vary considerably betweert countries and hence there is often a neeo i~ adjust the models to suit local conditions. Despite the importance of freight tariffs 4.3 The explanation of tariffs per Ifrm and Wlr km 9 5. The affect of roughness on revenues for differanf :outes 10 6. A mmparfson of date from rough and smooth mutes with vehicle operating cost model pradictfons 11 in Influencing the nature and location of economic activity Ji)d in determining the choice of mode and vehicle type relatively little analysis of freight tariffs has been carried out. It Is sometimes claimed that tariffs are unreliable and difficult to derive, but this was not found to be the case in Pakistan. Good quality, consistent data was readily obtained. These showed that from 1982 to 1986 tariffs rose closelv in tine with Inflation, The effects of flow Observed revenue and cost data were compared with estimates derived from the three standard models of vehlcie operating costs (VOCS) used for road appraisal in det’eloplng countries. Measured on a cost per kilometre basis, and excluding fuel consumption, both the Transport Research Laboratory’s Road Transport Investment model (RTfM), derived from data from Kenyan and the Caribbean, and the Worfd BanKs Highway Design and Maintenance Standards Model (HDM), using BrazilIan derived VOC relationships, were found to substantially overestlmata the costs and ravenues of trucks operating in Pakistan. The HDM VOC model incorporating Indian derived relationships, was found to give faldy close results for smooth reads; however In common wilh the other models, it also appeared to overestimate coats and tariffs for ve~ rough roads. Hence there were good grounds to befievg that all three models would overesti. mate the reductions in the per kilometre vehicle operating costs from Impmvlng unsurfaced roads in Pakistan. However corrections would also be needed to take account of the greater loads carrfed by vehicles on smooth roads. , 7. Trends in utifisation, ravanue and operating Costswith vehicle age imbalance: and seasonal variations could be easily seen. There Isa nb! outflow of goods by road from Karachi and as a result tariffs per kilometre were about onc third higher for vehicles traveling away fmm Karachi than in the opposite direction. A small but distinct seasonal pattern In tadffs was found. For journeys away from Karachi tariffs were above the mean by about 0.75 per cent for the first half of the year and they were below the mean by about 1 percent during the fatter half. The pattern of tariffs for journeys In the opposite direction followed an almost mirror image of this with lower tariffs In the first half and higher tariffs in the latter half of the year. 14 14 16 16 18 18 19 19 20 7.1 Annual distanm travelled 7.2 Revenua 7.3 Repair and maintenancecosts 8. Vehicfe profitablfity end trip distancg 9. Summary 10.Acknowledgements 11. Referenwa ,, Appendix The effects of distance, time, travel direction, load weight, road roughness and vehicle type on tadffs were investigated through the use of regression techniques. It was found that tariffs were much more dependent on journey distance than on journey time. Small and modeiate changes in road roughness had no significant effect. The following log-log multiple regression was estimated for the main paved road network, (N =1388, R’= 0.86): At the time of the surveys two-axle Bedford trucks with a seven tonne design capacity made up three quarters of the truck fleet, the remainder were mostly two and threeexfe Japanese trucks. The analysis showed that for long distance trfps the Bedford trucks were unprofitable. This was not the case for the Japanese trucks which have a larger mrrying capacity. Repair costs Increased, and annual travef decreased with vehicle age, but age amounted for only a relatively small proportion of the variation found In these items. It was estimated that after ten years, annual travel declined by 42 per cent for the two-axle Japanese trucks but by only 13 per cent for the Bedfords. I ,’ ,, ,’ ,., ; ,,, II ,, 1 ,. .... .J,. ..-.,. il t ~,,, ,’,2 ,, I ,,, , ,’ .,“1 ;,.’ PAKISTAN ROAD FREIGHT l~Du~~~~ AN A~A~~~~~ ; . “ OF TARIFFS, REVENUES AND COSTS ,, ABSTRACT A study of Pakistan’s private road freight Irenspofl industry was undertaken and data on freight tariffs, vehicle utilisation, revenues andoperating costs were analyzed, Over a period of four years it was found that average !afiff levels closely followed the rate of inflation with a small marked seasonal variation depending on trip direction. Using multiple regression analysis of crosssectional data, tariffs were shown to be a function of trip distance, trip time, direction, vehicle type and load weight. Only in the comparison of extreme conditions was roughness found to provide some additional explanation of freight tariffs. Oosewed revenues and costs were compared ‘withpredictions of vehicle operating costs given by the World Banks Highway Design and Maintenance Standards Model (HDM) and by the Transport Research LaboratoV’s Road Transport lnvestm6nt Model (RTIM). It was found that, in Pdkistan’s conditions, the models would over-estimate the reduction in the perkilometre vehicle oparating costs from improving the road network. Howaver allowances would also have to be mada to !ake account of the greater loads carried on smooth roads. An analysis of trends with vehicle &ge l~~dicatedthat annual travel fell with increasing vehicla age while maintenance costs rose. An examination of vehicle profitability showad that two-axle Bedford trucks (with 7 tonne load design capacity) were likely to be far more profitable on short distance routes (below 700 kms) while trucks with larger capacity were likely to be more profitable on longer distanca routes. The report supplements TRRL Resaarch Reports314 and VJ3, 1. INTRODUCTION This Report forms part of a study of Pakistan’s freight transport industry that was carried out by the Overseas Unit of the Transpoti and Road Research Laboratory in cooperation with the National Transport Research Centre of Pakistan In Islamabad. The overall alms of the study were to investigate the operation, performance, costs, tatiffs and management of the industry. The relevant information was collected during 1985 and 1986 by formal suweys and a number of informal Inlawiews and meetings. A general description of the Industry is provided in Research Report 314 and an analysis of vehicle time utilisation Is provided in Research Report 333, This Repoti addresses three separate, but Interrelated, Issues of national transport planning. These are:- a) Freight tariffs play a key role In Influencing the nature end location of economic activity aid they are critically important in both the choice of mode and the choice of vehicle used to transpod freight. For planning purposes it Is Important to bo able predict freight tariffs as accurately as possible and b) c) 2. for this reason it is useful to try to model fleight tariffs directly. Information on vehicle operaling costs can cedainly assist with this task but it is often unavailable and additional information on profitability, load factors, and empty running is also required. In order to assess the efficiency and long term viablli~ of the different components of Patistan’s road freight transport Industry it is Impotiant to be able to quantify total revenues and operating costs and hence identify the overall profitability of the indust~ and how the compocltion of the industry may change over time. There is a naed to improve our modelling of vehicle operating costs (VOCS) for mad Investment planning. Standard models of vehicle operating costs have been developed for the appraisal of road investment in developing countries, However, in practice, it has been found lhat thare are a number of problems associated with the use of Ihese models. The form of the relationships betwaen the relevant physical and economic parameters and VOCS varies between countries, and the models often need 10be adjusted to suit local conditions. An analysis of revenues and tariffs can assist with modelling the most appropriate relationships, DATA SOURCES The Reporf draws on the following sources of information:- 1)The Rondslde Interview Survey This was the main data collection exercise of the whole study, In total 3500 truck drivers were intewiawed at 39 sites throughout Pakistan during the period from January to April 1986. Drivers were asked to provide details of their current journey and their previous empty journey (i.e. origin, destination, journey distance, journey time, load, and tariff). In addition they were asked to estimate their consumption of fuel, tyres and their expenditure on repairs and the average distance travelled. Other information gathered included data on vehicle age, make, type, value, ownership, fleet management, finance, total revenue, accldants and insurance, In the survey, tw~axle Bedford trucks accounted for 7ti per cent of all vehicles surveyed, while twmax!e Japanese trucks (predominantly Hino, ISUZUand Nissan vehicles) accountad for a further 14 percent of vehicles. Technical datalls of the different vehicles suweyed are Included In Tablo Al of the Appendix, further details are given In TRL Research Report 314. 3 psfiicular journey, wilh a pStiCUlar Commodity CS~~ by a ~tiwlar vehicle type. The tariff data were recotiad twice each month II~ostlyrevering the period fmm lg~ tO 1986. II) Truck &*m’ -t and Revenue Dlarlo9 ..-’ 5500 -—— I I I I 1982 I9a3 1984 1985 19 Fig. 1 Tariff trends - from Karachi, January 1982 to June 1986 Manydrfvam wsm foundto keep detailed -rds of heir uts and revenues.In total, Uanes of over 50 different trucks ware collected coverf!lg about 600 veticle months. The ds~ COllSCtadrelates to different pariods fmm 1974 to 1986. The data were aggregated intc monthly periods bafom being analyzed and to be rmnsislentwith the other survgys adjustments wem made for inflation; all Dia~ data are expressed at’1 986 prf~ws. :.,.,.. fn aff, 123 diffamnt tariff SetS were COll@Ctd. of these, 56 were collected !mm agen!s in Karachi, 29 from Lahore, 24 from Abbottabad, 11 {mm Quet:a and three from Rawalpindi. Tha vehicl~ types included two-axle Bedfo~ tmcks, twoaxle fsuzus and Hines, and tractor trailers manufactured by Nissan and Mercedas. 5000 Ill) Vehfcfe Activity Survey In order to identify seasonal patterns, any tariff sets which were fixed for a period of six months (12 periods) or more were excluded from alf analyses. This left a basic data set of 86 different tariff sets which ran fmm the beginning of January in 1962 to the end of June 1986, a total of 106 periods. Time series analysis was used to astabfish what patterns if any were present in the data. Simple regression lines of tariffs against time were also plotted through the data sets in an aitempt to identify past trends and seasonality. 3.2 GENERAL TARIFF TRENDS A regression line was fitted through the data, and from thS it WE3 found Ihat there was a nomina[ increase in tariff Ievefs of approximately 33 pgr cent from the pedod Janu~,ry 1982 to June 1966. This represents a real incre~]seof between zero and one per cent as the cons!~merprice Index rose by 32 per cent over the same period. Tyre prices also rose by 28 per cent in nominal term!}, fuel prices rose by 29 per cent and the price of a Bedford chassis mse by 18 percent. A clearly distinguishable seasonal pattern was also found in the data with a broad seasonal deviation from the annual mean of approximately PIUSone per cent in the first half of the year to -1.5 per cent in the latter half of the year. 3.3 TARIFF TRENDS BY DIRECTION In this suwey, data was mllected by cooperative drivers and survey staff specifically recruitedto travel with Ihe vehicles’ crew. A mntinuous remrd of each vehicle’s activities and the mts and revenuesIncurred wem recorded over periods lasting fmm one to four weeks. In total, &/periods of data wemcollected. Data was collected fmm July 1985 to September 1966. iv) Past Tarfff Data In order to identify trandg and seasonatity in tariffs, past tariff datawem collected fmm remrds of freight agents, In total 123dffemnl sets of tariffs were collected, each relatingto a particularjourney, cargo end vefrfcle type. v) Road Rougfsrass Date 4000 6 fn orderto assist with the analysis of COS!and tariff dafa, informationon mad roughness was required. To supplementdata already collected for most of the main roads, an additional survey of mad roughness was undertaken for thestudy. This survey primarily covered the rough unsurfaced roads of Baluchlstan and the more important interior roads of the Punjab. Data was collected on 70 mad hnkscovering a distance of over 5700 km, In the survey, roughness fevels were measured with a bump integrator unit attached to the back axle of a survey vehicle and the data was then cafibmted with a ‘MERLfN’, TRRL’s fow cost roughness measuring instrument (Cundill, 1991). In total, the two sources provided data on mad roughness for 176 links coverfnq 14,000 km.Alf 1.5 1.0 ,, ,,’, ,, ,! ‘, ,, .; ;, Tariff rates per kilometre, are markedly higher. throughout Pakistan, iur vehicfes traveling away from Karachi than in the opposite dlmctlon, ‘This difference reflects the higher flows of freight traveling inland from Karachi. In order to analyse the differences In tariff trends the data set was split by direction of travel; 73 tariff sets related to journeys made in the ‘from Karachi’ direction and 13 were in the opposite direction, ‘to Karachr. Mean tariffs did not increase significantly in mat terms for trips In either direction, roughness data in the Report are pre~ented in Bf units (i.e. in mtim). 3. TARiFF TRENDS AND SEASONALIW Before tariffs can be used as a basla for ggneral transport planning purposes It is useful to know how stable they are over the long tarm, and to what extent they might fluctuate fmm season to season. These issues are addressed in this Section. . For the data set, average tafiffs for trips in the ‘from Karachi’ direction were over three times the value of tariffs ‘to Karachi’. This difference predominantly reflects differences fn trfp length and vehicle typewithinthe data 8et, rather than the underlying Vadation in tariff rates due tO traffic direction. Figure 1 illustrates the trend of mean tariffs from Karachi’ over the four and a half year pedod and points towards a seasonal pattern. The seasonality is more clearly Illustrated in Figure 2 where the mean residuals of the best fit line for the ‘from Karach~ data have been plotted, This shows a pattern of deviation of -1,5 ● ~~ .~ M A MJ A so N D Fig. 2 Tariff seasonaflty - from Karachi, January to December 3.1 THE DATA I -2.0 I Past tarfff data was collected from freight agents located in five major ctmtras of Pakistan. The data consisted of past ramrds of the tariff levied by the fmlght agents for a ., 5 ,, . .... ..—. ,, ,. ..,, up to +1.1 percent In the first half of the year and down to-1.7 per cent in the se~nd half. I1 Figure 3 ahowathe trend for mean tariffs to Karachi’ and , indicates a stnkin91Ydifferent Patiem of seasonali~. When the maan o~~heresiduals of the best fit line for the ‘to Karachi’ data is plotted (Figure 4), the seasonal pattern appears almosl to be a mirror image of the pattern ‘Irom Karach~.Tariffs are relatively low during the first half of the year, between 3.3 and 0.1 percent below the best fit line, and 0,4 to 2,2 par cent above the line during the last five months of the year. It is thought that the seasonal patterns obsewed above reflect in part the agricultural cycle in Pakistan and the movement of agricultural commodities as well as the seasonal patterns of imports. During the latter part of the year, large quantities of rfce are movad from the growing areas of Punjab and Slnd to the port of Karachi, Between 1982and 1984,approximately 85 percent of all rice for exportwas shipped ~o KarechYbalween the months of October and January (Majeed, 1985). A similar pattern can be obsewed for the movements of cotton ‘to KaracW for expoti (Japan International Cooperation Agency, 1983).This Is Ilkely to incre?se demand for freight transpofl in this direction and therefore inflate tariff rates during this period as shown In Figure 4. Conversely,national statistics on commodities imported Into Pakistan between 1983 and 1985 (i.e. goods moving in the YmmKamch~direction) indicate that Imports are relatively higher durfng January to June and lower during July to December (Federal Bureau of Statistics, 1986). This would tend to increase tariffs in the ‘from Karachi’ dheclion during January to June; it would also tend to reduce taritis in the opposite direction during this time as there would be relatively fewer return loads available In the ‘to KarachYdirection for the Increased number of trucks t~lng to return. This pattern would appear to be in sympathy with the seasonality Illustrated in Figures 2 and 4. J ,’0 .,~ ., .:. 4. ACROSS-SECTIONAL ANALYSIS OF TARIFFS AND OTHER DATA fn the previous section freight tariffs were analyzed from a historical perspective to identify trends and sea~unal differences. In contrast In this section a cross-sectional regression analysis of 2 idnge c! data (trip distance, time, direction, load weight, vehicle type and raughness) is undertaken to explain freight tariffs. Because the data were collected over a comparatively timited period (i.e. four months) the Influence of time trends was not addressed 111this pan of the analysis. 4.1 THE DATA The main data sources for this analysia were the floadslde Interview Survey and the Road Ro~ghness Suweys. In order to Investigate the relationships between tariffs and road roughness, it was necessary to derive average road roughness values for each vehicle trip, This was computed by averaging the roughness values of all the links In the trip, based on the assumption that truck drivers chose the minimum distance path between their origin and destination. A variety of truck configuration and body types are utilised in Pakistan’s road freight indust~. As these each have different operating characteristics which alfect the tariff they levy, they were separated out into four groups (see Table 1). Freight rates for commodities transpofled by tanker (such as oil and petrol) are usually fixed, and so all tankers were excluded from the analysis. Vehicles carrying loads of less than one ton were also excluded. As before the data sets were split by direction of travel. Analysis of the data indicated that for journeys made on the unpaved network of roads in the Mekran desert area of Baluchistan a high road roughness level was found to be associated with high teriffs, a marked Imbalance in traffic flow direction and a high degree of empty running. To provide a better explanation of thd relationships that govern freight tatiffs on the major pa,t of the paved road network the Mekmn data were mostly excluded from tha analysia in this section. A more comprehensive analysis of the Mekran data, incorporating the effect of more extreme road roughness levels, ia made in the comparison of rough and smooth routes in Section 5. In this Section the term ‘tatif~ relates to the actual sum of money paid for a particular loaded trip. Hence ‘tariff per kilometre’ relates to the tariff paid divided by the loaded distance travelled. Empty travel Is not explicitly analyzed in the Section. 4.2 THE EXPLANATION OF TARIFFS A variety of regression models were used to analyse the data In order to provide tho best explanation of freight tariffa. The best overall explanation of freight tariffs was given by the log-log multiple regression shown in Table 2, This is highly significant with a high R2value, In the regression, tariffs are shown to be statistically related to direction, (represented by a “dummy variable”), distance, time, load weight and vehicle type (represented by new vehicle value). The term relating to direction ( e“ ‘~’~) Indicates that traffic traveling from Karachi will be 27 per cent tigher than that traveling in the opposite direction, The coafficlents for the last four variables are equivalent to elasticities between the variable and the freight tariff, Hence, for example, a one per cent increase in load weight will increase tariffs by 0.28 per cent. In the form of the regression model used here the derived regression line passea through the origin. TOtest whether there was an ‘irreducible minimum’ component of the tariff to account for the inescapable costs of undertaking and organizing the tdp a variety of constant sums of money, representing this component, were subtracted from the tariff to see whether the explanation of the regression model could be Improved. In the event, it was Folmdthat no Improvement could be made to the explanation of the basic model. —. .I .,..... .. ....—- 1982 1983 1984 1985 1986 Fig. 3 Tariff trends -to Karachi, January 1982 to June 1986 I ““c8z I 4,0 I I I I I I I I I I I I I I I I I t ) 1 I t I I J ~ M A L! JJ P s OND Fig. 4 Tariff seasonality - to Karachi, January to December 7 .. ,, ,/”” ‘“ !~, -.., --=. —-,, Mean values f~ythe data sets (1986 prim, excluding Mekmn data) Tw~’ Kxle Bedfords Two-Axle Japanese TOKarachi ?mm Karachi To Karatit Fmm Karachi N=494 N = ZY6 N=88 N=109 ,- — Distance (krns) — 665 616 925 859 Trip time (hre) 28.1 26.8 46.6 Cargo weight (tonnes) 40.0 7.9 6.3 12.3 12.6 Roughness (mtim) 4903 5158 5802 5415 Tariff (Rs) 1626 2153 3476 3956 Tariff ~r km. (Rs) 3.62 4.23 4.28 4.81 Tariff pur tkm.(Rs) 0.59 0.63 0.37 0.46 New whide value (Rs) 155W0 155000 337000 337000 Three-Axle tmcks Tractor Trailers To Karscni Fmm Karachi YL Karachi N=35 From Karachi N=46 N=43 N=37 1- Distance (kms) 965 1028 932 860 Trip time (hrs) 49.3 57,9 52.5 58.5 Cargo weight (tonnes) 18.5 24.2 ?4.0 31.5 Roughness (mm) 4497 4929 $s355 4649 Tariff (Rs) 3931 7113 4!17 8357 Tariff per km. (Rs) 4.63 7.06 4.7.t 10.09 Tariff per tkm.(Rs) 0.26 0.30 0.21 0.35 New vehicle value (Rs) 4WOO0 483000 534000 534000 Soum Roadside Interview Survey TABLE 2 The prediction of freight tariffs using a variety of vadables (Excludng Mekran data) St. error Ln Tariff= 1.7365 + 0.24 Direction Variable 0.0191 (to Karachl=O, fmm Karachi =1) + 0.4468 Ln Distance (km) 0.0264 + 0.1976 Ln Time (hrs) 0.0233 + 0.2822 Ln Load Weight (tonnes) 0.0218 + 0.3167 Ln New Vehicle Value (1000 Rs) 0.0267 , (Road Roughness not significant) N=1386 Rz= 0.839 t value= 37.9 . .... I ,’ ,, 8’ ., ( ,! ,. 1 ! [ I . I As e-ed, the a-s found that journey time and distance were tighfy mrrafated. Despite tis, the elasfitify between tariffs and journey distan~ was mnsistenuy found to be more than double that between tariffs and journey time !SW Tables 2 @ 3). The results mn be Qmpamd with the analysis of the vetide utifi~tion survey date given in Research Report.?&. In Ihat analysis it was suggestmf that journey disfsnm muld account for 55 per cent, and journey time 45 pr ce!lt, of the total influence that these two variables had on the tariffs of B-ford tmcks. The differences in the results shows the difficulties in distinguishing the separate effects of time and distance b-use of the strong correlation between the variables, In the analysis, an etasticify inefficient of 0.05 was found between road roughness and tariffs. However it was not statistically significant at the 5 per cent level of probability (t value: 1.16) and was omitted from the regression model. A similar regression was carried out with the Mekran data included (see Tabl~ A2 in the Appendix). In this regression road roughness was found to be significantly correlated with tariffs (t value: 5.14). However there are grounds to suggest that the elasticity coefficient (0.10) is biased because of the imbalance in traffic flows ond high level of empty running in the ~iekran. In Table 3, multiple regressions for the ir,dividual Vehicle ti~pesare presented. The R2values for these regress, )ns, and their associated t values, are less toan for the regression of tho comMned data showl in Tabfe 2. As before, road roughness was not found to be statistically significant. The coefficients relating to direction are higher for the larger vehicles. In pafl, th,ismaybe an effect resulting from the longer trips ma~s by these vehicles. The elasticities for load weight are higher for the interme- -4- + ~___=>J ___ . . . . . . ,, date sired vehicles (the two and thre-tie Japanese n tmcks) than br the *axle BeMord trudrs and the trector-tila~ ~S may reflect the wider range of loads earned by these vehicles. 4.3 THE EXPLANATION OF TARIFFS PER TKM AND PER KM In many situations, tariffs are expressed on a per tonne kilometre (tkm) or per kilometre (km) basis. Two further sets of analyses have therefore been cenied out similar to those desctibed above. Table 4 presents two multiple regressions which explain fariffs per tkm. 9ecause the first regression is mathematically closely related 10the regression shown in Table 2 many of the coefficients are the same. As expected, because of the emnomies of scale, the elasticities between tariffs per tkm and distance and load weight are negative. The R2value and t value are less than for the overall regression explaining tariffs. In order to help provide an estimate of tariff per tkm without knowledge of the journey time or the vehicle type the second regression was included in Table 4. Table A3 in Ihe Appendix presents multiple regressions of tariffs per km for the four different vehicle types. These regressions are closely related 10those shown in Table 3 and, as before, many of Ihe coefficients are the same but in this ca~\ the distance coefficients are negative. The R2 values and significance of the regressions are less than the corresponding repressions in Table 3. TABLE 3 The prediction of freight tatiffs for different vehicle types (Excluding Mekrafi ,latfl) Variable to Two-Axle Two-W,,e Three-Axle Tractor predict Ln Tariff bedfords Japar.ese JaFaneso Tmitem Constant 3.5488 ~!.3303 2.0328 3.3027 Direction .2211 (.021) Ln Distance .4349 (.029) Ln Time ,1984 (.026) Ln Load Weight .2165 (.024) .1287 (.056) .5692 (.084) .1604 (.07) .6227 (.063) .3122 .6645 (.066) (,082) .6661 ,4648 (.062) (.126) not sig. .2392 (.092) .5811 ,2761 (.097) (.117) N 1030 197 81 R2 80 .609 .786 t value .752 .716 32.9 13.3 8.8 6.9 (Note: standard ermre In brackets, all units as In Tablo 2) 1 TABLE 4 The prediction of freight tariffs per tonne km (Excluding Mekran data) Ln Tariff per tkm = 1.7365 St. error + 0.24 Direction Variable 0.0191 (to Karachi=O, from Karachi =1) -0.5532 Ln Distance (km) 0.0264 + 0.1976 Ln ~me (hrs) 0.0233 -07178 Ln Load Weight (tonnes) 0.0218 + 0.3167 Ln New Vehicle Vallle (1OOOR~) 0.0287 N= 1388 R’= 0.706 t value = 25.8 Excluding vehicle value and journey time Ln Tariff per Ikm 2.2672 St. error = + 0.2339 Cireclion Variable (to Karachl=O, from Karachi =1) 0.0205 -0.3342 Ln Distance (km) 0.0109 -0.5449 Ln Load Weight (tonnes) 0.0187 N= 1388 R’= 0.659 t value= 29.9 5. THE EFFECT OF ROUGHNESS ON REVENUES FOR DIFFERENT ROUTES The nationwide regression analysis prasentad in Section 4 has shown that small to moderate differences in roughness levels hava little apparent effect on tariffs. The addition of survey data from the Mekran region of Baluchistan, where the roads have ver] high roughness levels, does provide a significant statistical relationship between road roughness and tariffs. (See Table AZ in the Appendix). l!tlfodunalely howevar, the Imbalance in traffic flows and high level of emply running recorded in the Mekran lead to serious bias in Ibis analysis, To try to get over the problem, a comparison was made between the average expected revenu6s for two standardised round tdps (which includa empty running), one located in Baluchist6n and one located on the rest of network. The Baluchistan route was between Karachi and Turbat (in the Mekran) which had an estimated average roughness of 12900 m~m, measured on the BI scale. The other was synthesised from data derived from vehicles operating over comparable journey distances on the rest of the network in Pakistan (average roughness 4800 mtikm). Sixty eight vehicle trips were recorded in the Roadside Interview Sumey along the rough route, 54 of which were Ioadad. Of these, 15 trips were by two-axle Bedford trucks and 39 trips were by two-axle Japanese trucks. ,! 10 , The majority of the trucks surveyed were based in the Mekran region which suggests that they plied this route regularly, Estimated trip distances for this journey varied between 650 and 870 kilome[res. Forjourneys throughout tha rest of Pakistan of between 500 and 1000 kilometres in Ienglh, trip data was collected for 330 Japanese and Bedford two-axle trucks.Within this catego~, 294 journeys were loaded and of these, 237 were two-axle Bedfords, and 57 were two-axle Japanese trucks. Table 5 gives a breakdown of the data used in the analysis. From the data in Table o mean values of the expected revenue per kilometre &ld Ihe expected revenue per tonne kilometre were calculated for a standardised complete round trip foreach vehicle type. lntheestimation of revenue per kilometre, empty running was taken inlo account using the loaded ratios found in the survey. The results are shown in Table 6. Here and in the rest ot !he report the term ‘revenue’ Is used when tariffs are added together to produce a statistic. ‘Revenue per kilometre’ refers 10the revanua earned in a period divided by the total (loaded and empty) journey distance travelled, Table 6 shows that revenue per kilometre was recorded to be 10 per cent and 24 per cent higher on the rough routes than on the smooth routes for Bedford and Japanese trucks respectively. Beause of the lower average loads on rough routes the corresponding figures for tariffs per tonne kilometre are 63 per cent and 79 per cent higher, TABLE 5 Truck Trip Data (1966 prices) Route To/From Karachi Rough Smooth To From To From Two-axle Bedford trucks:. Loaded trucks 3 12 Empty trucks 105 5 132 0 Loaded ratio per cent 14 37.5 8 100 Mean loaded distance Km 88.2 781 94.3 726 Mean empty distance km 726 627 698 Mean load 719 751 tonnes 5.5 8.3 Mean taritf Rs. 8.0 900 4617 8.7 1980 2680 Two-axle Japanese trucks:. Loaded trucks 8 31 Empty trucks 15 9 42 0 Loaded ratio per cent 12 47.1 2 100 Mean loaded distance km 55.5 876 95.5 649 Mean empty distance km 748 676 847 Mean load 667 704 tonnes 7,2 9.4 11 13 Mean Iatiff Rs. 2825 5597 3014 4381 — TABLE 6 Expected Revenue Per Kilometre For Rough And Smooth Roads (1966 prices) Two-Axle Bedford Road Type Two-Axle Japanese Rough Smooth Rough Smooth Road roughness, (B1scale) mm/km 12900 4800 12900 4800 Standardised distancc km 754 713 763 798 Expected load tonnes 5.2 7>7 6.4 9.3 Expected revenue Rs 2477 2136 3464 2930 Expected rev/km Rs 3,29 3.0 4.54 3,67 Expected rev/[onnekm RS 0.637 0.390 0,712 0.397 6. A COMPARISON OF DATA FOR (Cundill, 1993). Vehicle operating cost model relalionROUGH AND SMOOTH ships were also derived from studies In Brazil and lndla. These are included as separate VOC options In the ROUTES WITH VEHICLE World BanKs Highway Design and Maintenance Model HDM3. The particular model relationships used here are OPERATING COST MODEL derived from Chesher and Harrison (1987), PREDICTIONS From the data collected in the main surveys, estimates of The data for the rough and smooth routes given In total lifetime operating costs for vehicles operating on the Section 5 were used in a comparison with predictions of main roads of Pakistan are given In Table 7. costs derived from three sets of vehicle operating cost Key input assumptions ralaling to the VOC models are (VOC) models used for road investment appraisal. The included in Table 8. To be consistent with revenue data, VOC relationships derived from studies in Kenya and the Caribbean by the Overseas Unit, TRRL are Incorporated market prices (including taxes and duties) are used for all items. The VOC models predict maintenance pafis costs into the Road Transport Investment Model, RTIM3 as a percentage of Ihe new vehicle price where both 11 !-. .-,, . . . . ., TABLE 7 parts and vehicles are priced without tax. Bemuse the avarega tax rates on vehiclas and pafls were balievad to be broadly similar no adjustments were made to equate tha pradicted maintenance pafls costs to market price Ievals, fual consumption between the medals relates to this. Baceuse of the lack of reliable information on gradients it was decided to omit fuel consumption from the comparison. ! Estimatad Lifetima Operating Costs Per Kilometre (1986 prices) Two-Axle Two-Axle Three-Axle Bedford Japanese Japanese : ,. :. Naariy all the main paved roads In Pakistan are flat and straight. The Mekran-Karachi road passes through two ranges of hills and naturally some gradients and cuwaIure are present. However calculations show that model astimates of spare parts and tyra consumption would be only slightly affected by the inclusion of thase gradients and curvature and so their effect can be ignored. Data on gradents, curvature and superalevation ware not collacted; road roughness was the only key diffarance betwaen roug}l ~nd smooth roads examined by the models. Apart from the key input assumptions, advisory “default” options were usad Distance per day Kms 329 304 373 Running costs Rs par km Fuel Crew Malntananca and repairs Tyres Oil and greasa Loading Iabour Octrol, police, taxes Agents commission 1,257 0,426 0,322 0.142 0.141 0.079 0.171 0.078 1.333 0.472 0.294 0.142 0.190 o,oes o.193 0.115 1.732 0.587 o,38e 0,213 0.193 0.149 0.183 0.110 The resulls of tha model predictions are shown in Table 9. The table shows big differences between the model estimates of fuel consumption and maintenance costs. In the RTfM relationships, fuel consumption is very sensitive to gradient and clearly a major part of the difference in Fuel consumption data collected from tho roadside interviaw suwey on the rough and smooth routes are given In Table 10. These data are deducted from the revenues shown in Table 6 to give tha obsewed Total nmnlng costs 2.616 2.825 3.555 TABLE 9 Estimated capilaf costs 0.243 0.365 0.429 Vehicle Operating Cost Predictions (Rs per Km, 19e6 prices) Two-Axle Bedford Two-Axle Japanese Road Type Rough Smooth Rouoh Smooth Nat Profit 0.060 0.157 0.396 Total Revenue per km 2.939 3.347 4,382 RTIM: Fual Oil Tyres Maintenance parts Maintenance Iabour Crew Depreciation Interest Overheads 0.612 0.107 0.443 4,2?0 0.310 0.494 0.195 0.121 1.310 o,5el 0.054 0.396 2.444 0.182 0.277 0.110 o.06e 0.622 0.597 0.107 0,493 4.057 0.245 o.4e6 0.652 0.347 ?.393 — 6.357 0.587 0.054 0.444 2.037 0.127 0.269 0.377 0.200 0.819 — 4.913 TAB1.E 8 Kay Assumptions For VOC Model Predictions (1986 pflcas) Two-Axle Bedfords Two-Axle Japanese All VOC Models:- Annual traval Naw vehicle price Tyre price Diesel pnca Oil price Crew cost Maintenance fabour Interest rate Gradient Curvature Engine power kms Rs Rs Rdltr 112,000 325,000 2,275 4.3 13.4 13 9.4 10 0 0 98 123,000 390,000 2,275 4,3 13,4 13 9.4 10 0 0 1eo RTIM total 7.663 4,934 HDM (Brazil): Fuel Oil Tyres Maintenance parts Maintenance Iabour Crew Depreciation Interest 1,344 0.071 0.353 4,033 0.233 0.376 0.303 0.:s2 1.345 0.053 0.341 1.717 0.150 0.210 o,ie5 0.111 — 4.112 1.499 0.071 0.371 3.450 0.196 0.372 0,356 0.214 — 6.528 1.618 0.053 0.364 1,469 0.126 0.192 0.203 0.122 Rfir Rdhr ?’0 tim “/0 HP RTIM Only:- Mean aga Years 9.3 Vehicle valudnew price “/0 3.4 Overhead rate 25 0/0 62 20 20 Gross weighl, rougtismooth Kgs 10600/13100 11800/14700 HDM (Brazil) total 6.895 4.146 HDM (India): Fuel 0.966 1,240 0.817 Oil 1.133 0.043 0.037 Tyres 0.043 0.037 0.397 0,293 0.360 0.272 Meintenanca parts 1.069 0.362 1.003 0.366 Maintenance Iabour 0.745 0.310 Crew 0.634 0.268 0.290 0.290 0.264 0,264 Depreclatlon 0.141 0.141 0.325 0.325 Interest 0.096 0.096 0.221 0.221 — — — HDM (India) total — 3.747 2.790 3.667 2,666 HDM Only:- Mean life Years Tare waight 12 Kgs 12 5400 7000 Mean load, rougtiamooth Kgs 5160U670 6368/925 12 13 m’ ,, TABLE 10 ,’ ‘Observed Fuel Consumption (1986 pdces) Tw&Arrle Bedford Tw&Axle Japanese Road Type Rough Smooth Rough Smooth Fuel consumption Rs per km 1.463 1.233 1.534 1.290 revenues, without fuel consumption, shown in Table 11. Table 11 also lists operating rests, after subtraction of fuel consumption, predicted by the different models. Table 11 demonstrates wide variations between the diffarent estimates of coste and revenues. As expected the Indian relationships appear to provide the best estimates of revanue~ the HDM (Bmll) and RTIM relationships predict costs that are up to four times the observed revenues. In Pakistan, revenues must cover conventional operating costs, profits and items such as loading and unloading Iabour, ocfroi charges (a local tax mllected on vehicle movements), freight agents’ fees and gratuities paid to police. Typically these additional costs account for about ten per cent of the total revenues earned. The analysis shows that all of thg models considerably over-predict the effect of Increasad road roughness on operetlng costs. This Is particularitymarked for RTIM and the Brtillan relationships, The tndian relationships for the two-de Japanese trucks provi~ed the closest prediction of the differences In tadff levels batween rough and smooth roads but even In this case the difference was overestimated by 43 per cant. Reductions in vehicle maintenance costs mmpdse most of the bnefita predcted by the models following reduced levels of road roughness. In mmparing the model predictions with astimated Iavala of maintenance costs dadved fmm empldcal data (see Table 12 In section 7,3) it would appear that this particular component is substantially ova~estlmated. ~le reasons for the low levels of vehicle malotenance costs in Pakistan are probably because of the widespread network of skilled mechanics, the availability of low cost, locally manufactured spare parts, and slow driving speeds. Another factor, which may also have an effect, is that oil changes appear to be far more frequent In Pakistan than In other countdes. A comparison of the Indian operating cost predictions with the predicted lifetime costs also shows Important differences in tyre consumption and crew costs. Tyre costs appear to ba overestimated whila crew costs are underestimated. Part of the explanation of the differences in tyre consumption may relate to tha slow running speeds and the widespread use of secondhand remoulded tyres in Pakistan. The differences in crew costs relate to the genarally higher veh!cle manning levels in Pakistan than in other countdes, 7. 7.1 TRENDS IN UTILISATION, REVENUE AND OPERATING COSTS WITH VEHICLE AGE ANNUAL DISTANCE TRAVELLED TABLE 11 Three estimates of annual distance travailed have been calculated from tha Roadside Interview Survey (See L ,,; I Compadson Between Observed Revenuee and Predicted Costs (Excluding L, ,;. fuel consumption component, Rs per kllomatre, 1966 pdces) ,tie” Two-Axle Bedford ~:. .: : Two-Axle Japanese Rough Smooth .:, Route Differ Route Rough Smooth [m, Differ -ence .#: .:/ Route Route -ence 1 \ ... ~,.,:. Obsejwed revfim i .803 ,1.766 +0.037 .,;,:!,. 3.007 2.362 +0,625 ; 1,,,,; !.>.,., RTIM costim 7.251 4.353 :g(~ t2,696 7.760 4.326 +3.434 1 ;~..,)..;:1 HDM (Broil) cos~m 5.551 2.767 +2,764 5.029 2.526 +2,501 ‘%Ng{ ,, .:) HDM (India) COSW 2.7aj 1.550 +1,231 2,a50 1.753 +1.097 .’,,’-7 .. ~,:;?:$r ,:~::~..! I .. Table A4 of the Appendix), For two-axle Bedford trucks, the mean of the estimatea given ia 112,000 kms per year. Further ganeral confhmation of these estimates was collected from the Vehicle Activity Suwey and the Drfvere’ Dlatia~ In the former, two-errle Badford trucks were estlmatad to travel 109,000 kms per year and In the latter 122,0C0kms per year. Using data from the Roadside Intewiaw Survey, an analyals was carried out to detennlne how annual travel Is influenced by vehicle age. Two estimates of annual travel were used for this analysis. Ol]e estimate waa . based on the drivers’ own estimatea of their normal weekly travel (adjusted for days off the road under repair) and a second estimate was derived from the current trips of loaded trucks using the time and distance taken for the current loaded journey and from any previous empty trip undertaken, (Both types of estimate were used to calculate the mean annual distance travelled shown In Table A4 discuaeed above.) The results of this analysis are shown In Tabla A5 of the Appendix. 50 ● o The regressions given In Table A5 show that vehicle age has a very significant effect on annual travel, although it accounts for a relatively small proportion of tha total vadatlon. When estimating their annual travel, drfvere tend to Imn out their daytoday fluctuations and so a closer ralationahip can be expectad between age and the ddvere’ own estimates of annual travel than between vehicle age and the estimates based on current trips. This is confirmed by the analysla. Vehicle age affacts the utilisation of the two-axle Japanese trucks far more than Bedford trucks. For example, by applying the regression equations for the data eeta derived from the drivers’ own estimates of annual travel, It can be ?celculatedthat after tan years, the dlstanca tmvelled per year will decline by 42 per cent for two-axla Japanese trucks and only 13 per cant for the Bedford trucks. Mean estimates of annual distance travelled for Bedford trucks and twoaxle JaDanese tmcks are given in Figure 5. ~ ~1 ..~l 2 4 6 b 10 ,T~— 20 Vehtclo ago (years) Fig. 5 Distance travelled per year 15 ., , ,. ,., .. ,, I I ... .,. .,, . ,.’ 7.2 REVENUE J ~~ ... .,, ... .. ,,, ,., , Dala on vehicle eemings was collected from tde Roadside irttetiew Sutiy; Drfvers’.Diarfes, and frnm the Vehicle Actfvfty Suwey. Estimates of earnings per day from the different surveys ISgiven in Research Report 314. The estimates of mean revenue per day for the twoaxle Bedford trucks range fmm 967 to 1086 Rs per da~ a difference of about 12 per cent. Based on information from the main Roadside Interview Survey, estimatwJ mean revenue per day for two-axle Japanese tn:cks was 1171 R%for Ihre&exle Japanese Imcks 1664 Rs; and Japanese Trector-Trsllere 1917 Rs. Bemuse of the smaller sample sizes, there la greater uncertainty concerning the eslimates of revenue per day for the larger tmcks. As a vehicle ages, it can be expected that dtstance travelled and revenue eamed,will decline. The regression analysis presented in Table A5 and discussed above found that annual distance trevelled did decline with vehicle age for Bedford and tw~axle Japanese tmcks. However using data fmm the Roadside Interview Survey no significant relationship was found between revenue per day and vehicle age for Bedford trucks. This Is principally explained by a rise in revenue per kilometre with age as o!der vehicles are diverted towards shorter mutes. Nevertheless revenue per day will almost certefrrly decline towards Ihe very end of a vehicle’s life as it becomes very unreliable, The oldest and least reliable vehicles tend to be used most on short distance urban mutes. These vehicles may not have been detected because Irrtra-dfstrfcttraffic was omitted fmm the Roadside Irrtewiew Survey. In contrast to the results for the Bedford trucks, revenue per day for the two and three-axle Japanese trucks was found to decline with vehicle age as annual distance travel fell. For these vehicles, it was calculated that revenue per day would decline by about one third after ten years. No significant relationship was found between age end revenue per kflometre, It is difficult to explain the differences in opemtlng performance found between the Bedford and the Japanese two and three-axle trucks, An analysis of revenue per kflometre and revenue per day was carrfed out using data for Japanese tractortreiler vehicles. No significant relationships were found. 7.3 REPAIR AND MAINTENANCE COSTS It is generally accepted that vehicle repair costs will vary with vehicle aga and mad condition. A summary of data collacted on repair costs is shown in Table 12. As TA5LE 12 Geneml Repair and Malntananca Costs (Rs par km, 1966 prfces) Two-Axle Three-Axle Tractor-Trailers Badfords Japanese Japanase Japanase ‘~ mean 0.239 0.2M 0.315 St. Error 0.015 0.03 0.04 ( mean vehicle age 4 yre 1 yr 1 yr Excluding trucks operating on ,,.i .1 7 rough or mountainous terrain !,,., ,.,,;, mean :., St. Error (!* :( .>’ .. mean vehicle age Trucks opemtlng on mountainous nodhem mutes mean St. Error mean vehicle age $J Tmcks opemtlng on J rough reads in Mekmn .:i: mean g: St. Error ,,, mean vohlcle age 0,358 0.259 0.367 0.005 0.009 0.036 loym 3 yre 3 yrs 0.568 0.043 4 yrs , 0.573 0.025 llyrs 0.351 0.31 0.028 0.031 9 yrs 2 Vrs 16 ,,, , ... .,.,, ‘,, ,,, I expected, vehlcla repair cosls ware found to have a high Uegme of variability. Tha mean estimates repofled fmm tha Diarfas are lower than the estimates fmm the Roadside IntervlelwSurvey. In pa:t this can ba explained by tha lower vehicle aga of the Diary Data. Badford trucks travailing on the mountainous northern mutas appaar to have higher repair costs than thos- *nicks trsvelfing on flattar terrain. The Japanese tnrck~ ,revelllng on the rough roads in the Mekran were also reported 10have higher repair costs although this was not the case for the Bedford trucks, Some of the differences in costs maybe obscured by differences In age structure; this Is examined in greater detail below, To determine the effect of vehicle age on repair costs a setias of regressions were mrrfed out using data from the Roadside Interview Survey the results are givan in Tabfe A6 in the Appendix. This analysfs showed that age can account for only a small proportion of the totaf variation in repair costs afthough for two-axle Bedford and Japanese trucks Ihe regressions relating repair costs per tilometre to vehicle age are significant. The repair costs (per km) of the two-axle Japanese trucks are shown to start at a lower base but to grow more quickly than the repair costs of the Bedford trucks. No significant trends coufd be 0.2 0.1 . .. .,:,,. :.. ..( ,., . . ..,.,.,. i’: ;..,.:-’ i., ,. iderrtffied between repair costs per day and vehfcfe ag~ :., ,,, ,,.,.:. this is probably because the higher repair rests per ;“’ .. . .. ....’ Nfometre are mmpensated by lower annuaf tmval as the ,., ., i.’ .:...” vahicfe ages. fn Ffgure 6 the mean repair rests (per km) ,., ‘! are plotted against vehlcfe age for the twwexle Bedford ,, ,,. , and Japanese trucks. ., The regressions given in Table A6 show that the repair costs of two-ssle Japanese tNcks operating on the rough roads in the Mekmn increase at a much faster mte than for those operating in the rest of the country. In contmst, no significant relationship coufd be found betwaen repair costs and vehicle age for the Badford trucks operating In the Mekran. Spare parts for the Badford tNck ware found !(Jbe widely .. availabfe all over Pakistan. In tile larger towns oll~!nal parts for the Japanese trucks wera also found to be available. Many parts for the Bedford are manufactured In Pakistan and in most ~sas If a part Is not at hand it can be made at one of the numerous small workshops. A small survey of vehicle spare parts was csrr!ed out in Rawalplndi and Lahore. It was found that Japanese tNck parts were on sale at about three times the price of the equivalent for a Bedford truck, ? I I 1 I I I I I I I 2 4 6 a 10 12 14 16 16 20 Vehlcloage(years) Fig. 6 Truck repair costs i. ~ ‘ ,. -,. 8. VkHICLE PROFlTAf31LlTY AND :. TRIP DISTANCE Anarrafyefsof vehicle profitability is repo~d in Research Report 314. Assuming that fhe,overafl revenues and costs will be maintsdnedin rest terms in the future for veticfes of different ages It was estimated that the tw~ axle Bedfod trucks were just profitable (IRRs in the mnge 6-9 par cent). Other truck type wem found to be more profitable than the Bedfords the three-axle Japanese trucks were found to be the most profitable of all (IRRs 5&70 percent). Using the date collected, an investigation was made on how pmfitebihty changes with tdp distance. Hgure 7 shows this relationship for twmexle Bedford trucks. The analysis indicated that for long distance trips the two end thre~axle Japanese trucks were profitable while the Bedford trucks were not. This confirms the widely held view that trucks with small ce~ing capacity are more suited to short distance journeys where their flexibility is an advantage. 9. SUMMARY To examine the effect of large differences in road roughnesson revenues, (and allowing for differences in empty running) a comparison was made between a very rough mute in the Mekran (13,000 mtim) and mutes on the rest of the main road network (4,800 mtim). On the rough route, revenues per vehicle kilometre were estimated to be 10 percent higher for 2-axle Bedfords and 24 per cent higher for 2-axle Japanese trucks. However because of the lower loads carried on the rough route, revenues per tonne km were 63 per cent and 79 per cent higher for the Bedfod and Japanese trucks respectively. 11. REFERENCES An analysis of trends in freight tariffs over the pedod fmm 1982 to 1986 showed that tariffs mse Ctosely in line with general price inflation. A small, but distinct, seasonal pattern in tariff levels was found. For journeys outbound fmm Karachi tariffs wero above the mean by about 0.75 per cent for the first half of the year and they were below the mean by just over 1 per cent during the latter half. The pattern of mfiffs for journeys in the opposite direction followed an almost a mirror image of this with lower tariffs in the fimt half and higher tariffs in the latter half of the year. Chesher A and RHardson(1967), Vehicle operating costs evidence from developing countdes. H;ghway Des;gn and Maintenance Standards Serias, John Hopkins University Press, Baltimore. Cundill M A (1991), The MERLIN low-cost road roughnessmeasuring machine. Department of Trarrspoti, TRRL Report RR 307. CroWhome ~ransport and Road Research Laboratory). CundillMA(1993), The road frenspoti investmarrt mode/ (RT/M3) usefls marruaf. Department of Transport, Transport Research Laborato~, Crowlhorne. Monthly Statistical Bulletin (Janua~ 1986), Federal Bureau Of Statistics, Statistics Division, Government of Pakistan, vol. 34 no.1. Karachi. ‘:~ ,, .I ,. 1. . The revenue data were then compared with estimates derived from standard models of VOCS used for road appraisal in developing countries. Measured on a cost per kilometre basis, and excluding fuel consumption, both the RTIM model and the HDM Brwil model of VOCS were found to substantially overestimate the costs (and tariffs) found for trucks operating in PaMstan. The HDhl Indian model was found to give fairly close results for smooth roads: however in common wilh the other modefs, it also appeared to overestimate costs and tariffs for very rough roads. Hence there were good grounds to believe that all three models would ovarasfimate the ,; The effects of distance, time, travel direction, load weight .. end vehicle type on tariffs was investigated through the use of regression techniques. A high degree of explanation of the variability In tariff levels was found using a Ioy log multiple regression (overall R2= 0.64). The effect of journey distance was consistently found to be much greater than the effects of journey time. Small and moderate changes in road roughness were found to Impose no significant effect on tariffs. Majeed, A. (1965), Foodgraln Transport Economics and Logistics Study, Government of Pakistan, Ministry of Food Agriculture and Cooperatives (Pakistan Foodgrein Storage Project). NTRC, Plannlng and Development Division, Islamabad, reductions in the per kilomelre vehicle operabng costs from improving unsurfaced roads in Pakistan, However corrections would also be needed to take account of the greater loads carried by vehicles on smooth roads. Repair costs were found to increase, and annual ~stance travelled was found to decrease with vehicle age, although age could account for only a relatively sm~;i proportion of the variation found In these items. For twoaxle Japanese trucks, revenue per day was found to decline with vehicle age although no similar trend wae observed for two-axle Bedford trucks. Indications were found to suggest that long distance ttips were unprofitable for the twc- axle Bedford trucks. In contrast long distance trips were found to be more profitable for the two and three-axle Japanese trucks which have a larger carving capacity than the standard Bedford truck. Japan International Cooperation Agency (1963), The Study on National Transport Plan in the Islamic Republic of Pakistan, Tecfrnicaf Paper Vo/ume /. Other Research Reports prepared by the Overseas Unit, Transport and Road Research Laborato~, relating to the Pakistan road freight indust~ research project are ae followsHine J L and A S Chllver (1991), Pakistan road freight indust~ An ovewiew. ~e~eflment of Transpod, TRRL Report RR 314, CroMhome (Transport and Road Research Laboratory), HineJL(1991), Pakistan road freight indust~ The productivity and time use of commercial vehicles. DepaRment of Transporf, TRRL Report RR 333, Crowthome (Tmnsport and Road Research Laboratory). 1300 ., 10. ACKNOWLEDGEMENTS The work desctibed In this report forms part of programme of joint research between the Overseas Resource Centre (Research Directoc Dr. J Rolt) of the Transport Research Laboratory, and the National Transport Reseerch Centre, Pakistan. The assistance of Dr. M Cundlll (Overseas Resource Centre) is also acknowledged. f Tflpdistanco(kms) ->-: Fig, 7 Revenueand costs of Bedford trucks ,: ,, .Y. ,. ‘.,, I i, ::18 ,, 19 I ... .. .. ,,: , i, :1 i- { i, APPENDIX TABLE Al Common Trucks In Pakistan Vehicle Model Type Axles GVW* GCW.. HP Price Rs. Make Kg Kg Sept 1986 Bedford CJP Rigid 2 10,920 nla 98 275,000 Bedford TM2500 Tr. Unit 2 tia 25,000 171 Hino da FF 170 Rigid 2 da da 200 412,000 Isuzu JCWFTR Rigid 2 12,000 nla 160 398,000 Isuzu TDJIDVR Rigid 2 15,000 27,000 220 515,000 Mitsubishi FP415ER Tr. Unit 2 15,400 39,000 310 730,000 Nissan TK20GT Tr. Unit 2 14,175 26,000 190 570,000 Nissan TK20 Rigid 2 16,500 26,000 190 475,000 Nissan TD1O Rigid 3 23,000 nla 160 480,000 Nissan U780E Rigid 2 12,000 nla 140 342,000 . Gross Vehicle Weight ..Gross Combination Weight Source: Manufacturers Speciflcatlons TABLE A3 The prediction of freight tariffs per vehicle kilometro for different vehicle types, excluding caaea with road roughness >10,000 mm/km Variable to predict: Two-Axle Two-Axle Three-axle Tractor Ln Tarifl/km Bedfords Japanese Japanese Trailers Constant 3.5468 2.3303 2.0328 3.3027 Direction .2211 ,1287 (,021) .3122 (.056) .6645 (.066) (.062) Ln Dlatance -.5651 -.4308 (029) -.3319 (.064) -.5352 (.062) (.126) Ln Time ,1984 ,1604 not sig. (.026) (!07) .2392 (4092) Ln Load Weight ,2165 ,5227 (.024) .5811 (.083) .2761 (,097) (,117) N 1030 R, 197 81 .564 60 .286 .625 t value 18.2 4.4 .652 6.5 5.9 TABLE A2 The prediction of freight tarfffs for different vehicle types, includng Mekran data St. error Ln Tariff = 0.787 + 0.2605 Direction Variable 0.0195 (to Karachi=O,from Karachi =1) + 0.4417 Ln Distance (km) 0.0268 + 0.1919 Ln Time (hrs) 0.0241 + 0.2606 Ln Load Weight (tons) 0.022 + 0,3426 Ln New Vehicle Value (1000 Rs) 0.0264 + 0.1845 Ln Road Roughnesst (Bl:mdkm) 0.0359 N=1451 R’= 0.632 t vafue = 34,5 ~, Note this coefficient Is considered unreliable because of the high degree of empty running on very rough roads (see main text). 20 ,m. --:, L—..-..—-— — (Note: standard errors In brackets, all units as In Table A6 ) TABLE A4 Three Estimates of Annual DistanceTravelled(1000 kms) Two-Axle Two-Axle Two-Axle Two-Axle Three-Axle Tractor-Trailer Bedford Hino Isuzu Nissan Nissan Nissan 1) based on weekly distance 117 159 147 132 143 136 ii) based on trip revenues” 109 116 104 95 112 129 Iii) based on trip times 109 129 117 106 120 127 Mean of estimates 112 135 125 112 125 131 Source: Roadside fntewiew Suwey . Excludes data from Suwey Slations 1-11. ,, 21 ,, .. ~,’,,,,,”: 1 ,, 1 ,1 :\, : TABLE A5 The Fekdlonship Between Annual Distance Travelled and Vehicle Age (In years) Two-Axle Bedford Trucks Kllometres per year= 141,629.1816 Vehicle age N=1848 (se= 170) R’= 0,058, t value= -10,7 Two-Axle Japanese Trucks Kilometres per year = 166,263-6923 Vehicle age N = 434 (se = 821) ~2= 0.14, t value= -8.44 Three-Axle Jepanese Trucks Kllometres per year= 154,977-2957 Vehicle age N=75 (se = 1610) R2 = 0,045, t value= -1,837 Source: Roadside Intewiew Suwey TABLE A6 The Relationship Between Repair Costs and Vehicle Age (Age in years, 1986 prices) Two-Axle Bedford Trucks Repair cost per km = 0.303+ 0.0056 Vehicle age N= 1977 (se= 0.0009) R’= 0,02, t value= 6.437 Two-Axle Japanese Trucks Repair cost per km (excluding trucks operatin~ on rough or mountainous routes) = 0.219+ 0.0166 Vehicle age N = 346 (se= 0,0037) R2= 0.06, t value= 4.57 Repair cost per km ( trucks operating on rough routes In the Mekran area) = 0.17+ 0.07 Vehicle age (se = 0,02) N=74 R2=0,15, t value= 3.53 Three-Axle Japanese Twcks Repair cost perkm =0.317+ 0.0195 Vehicle age (se = 0.015) N=107 R’= 0.016, t value= 1.3 Source: Roadalde Interview Survey Pdn!ti in thoUnllOdKlnQdo”l for HMSO 22 0dK03so0W4 C5 QM2 1017o . , MORE INl~oRMA7rIoN FROM TRL TRL has publishedthe following otherreports on this area of rcsemh: RR314 Ptistan road freight industry: An overview, J L Hine ~d A S ChiIver RR333 Ptistan road freight industry: The productivity and time use of commercial vehicles, J L Hine If you would like copies, photocopy and fill in the slip below. Enquiries to Overseas Centre, 0344770187 or 770167. To: Overseas Centre, Old Wokingham Road, Crowthome, Berkshire, RG 11 6AU. Pleilsesend me the following TRL reports (state report Nos and quantity) ........................................................................................................................................................... ........................................................................................................................................................... ..................................,...,..,.,,,,,,,,,,..,,...,.............................................................................................. ........................................................................................................................................................... Name .,,............................................................................................................................................. , Addres~. .............................................................................................................................................. ,,.,,,.,,...,,., ,,.,,,,.,,,...., ................................,,..............,,,,,,,,.,.,..., ........................................................... ........................................................................................................................................................... ........................................................................................................................................................... Postcode ..................................................................................................... ...................................... Telephone .......................................................................................................................................... .................................................,,............................................................................................. ........... ,,. Our 3W scienti~ ond en~ekrs ticlude mky world-cbtssex~rtr.ori Mghwqjsdc~& ad mtitenance, trwport structur~, trtilc systems,veMdeMely~~~d snfetymdthe” environment.