Effect of the modified bus segregation scheme (MBSS) on bus operation along the Epifanio De Los Santos Avenue (EDSA)

EFFECT OF THE MODIFIED BUS SEGREGATION SCHEME (MBSS) ON BUS OPERATION ALONG THE

EPIFANIO DE LOS SANTOS AVENUE (EDSA)

Mr. A le xis M . F illone

Civil Engineering D e p a rtm e n t, De La Salle U niversity

Abstract: The M odified Bus Segregation Scheme (MBSS) being implemented by the Metro Manila Development A u th o rity (MMDA) to improve bus operations along EDSA was analyzed by taking bus samples before and after its implementation. To determine its effect on bus operations, such bus operating characteristics as travel and stop times, average tra v e l and ru n n in g speeds, and average passenger-kilometer performance were used as measures.

Considering the segment o f EDSA from Gil Puyat Extension to Aurora Boulevard, the result showed that in general, the average travel and stop time have increased after the im plem entation o f the MBSS, although there was a noted decrease in stop time fo r a given value o f travel time. It was also known th a t the average passenger-kilometer performance o f the buses has n o t change after the im plem entation o f the MBSS although th e ir average travel and running speeds decreased. Also, no significant differences were n o te d in the o p era tin g perform ance o f the num bered buses, although in the volume survey buses numbered one(1) in operation were quite few compared to buses numbered tw o(2) and three(3).

1. INTRODUCTION

Public utility buses operating along EDSA are considered one of the main causes of traffic congestion in the area, especially near bus stops. Congestion n o rm a lly o c c u r w h e n th e y o u tm a n e u v e r e a c h o th e r to get passengers thereby blocking adjacent lanes. Bottlenecks and constriction of traffic lanes used by other vehicles result as manifested by long vehicular queues before major bus stops. In addition, the MRT construction which occupies at least tw o lanes (one in each direction) of EDSA has farther

reduced the available num ber of traffic lanes. For the time being, the lanes occupied by the MRT construction w ill rem ain w h ile th e p ro je c t is ongoing. Being a priority project, it is expected that once the MRT system is operational, it will greatly improve commuter travel along EDSA. It is then im p e ra tiv e to in tr o d u c e tra ffic m a n a g e m e n t m e a s u re s on bus o p e ra tio n s to alleviate th e traffic congestion.

Hence, the introduction of the bus s e g re g a tio n s c h e m e s (MBSS) to im prove bus operational behavior along EDSA. While there w ere oth er

tra ffic m a n a g e m e n t m e a su re s introduced before, such as the yellow bus lane and th e bus segregation scheme, there has been no schematic e v a lu a tio n o f w h e th e r th e n e w schem e is an im provem ent of the previous schem e. However, going over the details of the new schem e, buses will still be using the segregated yellow lanes while also being allowed to use the under- and over-passes w h e n e v e r p o ssib le an d th e tw o g ro u p in g o f th e p r e v io u s b us s e g re g a tio n s c h e m e h av e b e e n expanded to th ree bus groupings.

Further each grouping can only stop at 10 of the 22 designated bus stops along EDSA. This latter was done p rim a rily to a n s w e r th e b u s congestion problem near major bus stops. At the very least these are the noted changes in and improvements o f th e m o d ifie d bu s se g re g a tio n scheme.

This study aimed to determ ine the effect of the MBSS on bus operating c h a ra c te ristic s along EDSA. Bus samples w ere taken before the MBSS was introduced in November 1 2 , 1997 and additional samples w ere again taken after its im plem entation, since c o n g e stio n along EDSA is w o rse during the peak hours (morning and afternoon), bus data w ere collected during these periods on weekdays.

Data w ere gathered using the classical survey m ethods such as volume and boarding check surveys.

By concentrating the analysis on the Gil P u y at E x te n s io n to A u ro ra Boulevard segment of EDSA, being the segm ent used by all buses passing

EDSA even though their origin and d e s tin a tio n p o in ts are d iffe re n t, se v e ra l in te re s tin g re s u lts w e re o b ta in e d . For o n e , th e averag e passenger-kilometer perform ance of the buses has not changed after the implementation of the MBSS although th e ir average travel and ru n n in g speeds decreased. Also, significant d if f e r e n c e s w e re n o te d in th e o p e r a tin g p e r f o r m a n c e o f th e n u m b e re d buses, a lth o u g h buses num bered one ( l ) in operation are quite few com pared to bus num bers tw o (2) and th ree (3).

T his study, how ever lim ited , has evaluated one of the more economical m e a s u re s o f m a n a g in g tra ffic c o n g e s tio n th r o u g h p u b lic transportation system management. It is hoped that the current governm ent should put m ore em phasis on this approach to solve traffic congestion e s p e c ia lly w ith th e lim ite d infrastructure funds available.

2. T H E O R E T I C A L CONSIDERATION

Public bus journey time can be divided into three parts, namely: (1) Moving tim e; (2 ) tra ffic d e la y s s u c h as in te rse c tio n delays; delay d u e to interaction with other vehicles; delay c a u s e d by e m b a r k in g an d disem barking passengers; and other related delays; and (3) turnaround tim e. Variability on m oving tim e depends largely on the level of service of the road: the m ore vehicles the m ore c o n flic t occur, resu ltin g in slow er overall speed of th e traffic

stream. Since city buses operating in Metro Manila have to com pete w ith road space, it is expected that the incidence of traffic delay is high.

S e rv ic e r e lia b ility is a c o n c e p t frequently applied in measuring bus- operating performance. According to Jordan and Turnquist (1979), it reflects the degree to w hich buses provide regular or “on-time” service. Along th e s e g m e n t o f EDSA, s e rv ic e reliability is quite high; a bus arrives approximately around 7 to 11 seconds in any direction during the peak hour.

Hence, the problem of bus service p e r f o r m a n c e lie s m ore on th e variability of moving time and actual time spent on bus stops waiting for passengers.

Several p u b lic bus tra n sp o rta tio n system im provem ent techniques are available to improve bus operation in the urban area (Tanaboriboon 1992).

This in w ith -fl o w bus lanes, contra­

f l o w bus lanes, bus-only streets, bus ways, transit priority a t traffic signal, bus p riority f o r access to arterials by signal, a n d preferential treatm ent o f th e b u s e s o n fr e e w a y s . T h e s e te c h n iq u e s a re c u rr e n tly b e in g applied to im prove bus operation.

These w ere the following:

A. th e EDSA Bus Lane S c h e m e ( 1990) w hich designated tw o of the outermost lanes among the six lanes in EDSA as bus lanes on weekdays and yellow lane m ark­

ing was applied to separate the bus lanes from the ordinary lanes.

Initially, the bus lane demarcation was enforced during morning and afternoon peak hours but was

later changed to w hole day, ex­

cept on Saturdays, Sundays and holidays. However, confusion prevailed as to h ow th e measure was to be followed and enforced betw een the drivers and traffic enforcers. The problems encoun­

tered included the misconception that the lanes can confine buses w ithin and hence could n ot use the inner lanes and th e rules for right turning vehicles w ere not clearly defined. D espite these shortcom ings, th e m easure was officially hailed to have improved bus operations along EDSA; and

B. the Bus Stop Segregation Scheme w hich was intended to make the loading and unloading of passen­

gers at bus stops m ore organized by dividing th e bu ses plying EDSA into tw o groups based on their destinations. The bus stop area was delineated at different locations for each grou p w ith markings and signs indicating the group num ber. Buses w ere re­

quired to stop and load/unload passengers only at designated bus stops. The schem e was expected to enhance the convenience of public transportation.

The MBSS is the new version of the bus stop segregation schem e, and it was implem ented starting November 12, 1997. U nder the schem e, buses plying EDSA w ere divided into three groups, with each group allowed stop at no more than 10 of the 22 identified bus stop of EDSA from Mantrade to Balintawak. However, a bus may stop at any bus stop outside of the given

limits or boundaries. The groups were num bered 1, 2, 3, and assigned with colors Tangerine, Yellow-Gold and G re e n , re s p e c tiv e ly . O n e -fo o t re flec to riz ed stickers bearing the color-coded numbers, as manufactured and controlled by IMBOA (Integrated Metro Bus Operators Association), are placed at th e right w in dshield to identify the buses. The same color- coded num ber is painted on the left side of the door. IMBOA is responsible for assigning bus stops and num bers and issuing stickers to buses. Non- IMBOA members have to communicate directly with IMBOA for their bus stops assignments and stickers. Signs w ere installed (just like the first bus stop segregation schem e) at the different bus stops to indicate which buses stop there and help commuters identify the buses they have to take.

An initial study done by the author, (Fillone, et. al 1998) show ed that the average passenger-km performance of air-conditioned buses along the Gil Puyat Extension to Aurora Boulevard segment to EDSA did not change after the im plem entation although there w ere some noted decrease w hen the whole route system is considered. On the other hand, the average travel and running speeds decreased after the im p le m e n ta tio n a lth o u g h o th e r extraneous factors such as the MRT construction and the pre-Christmas season traffic could have caused it.

W ith a d d it io n a l i n d e p e n d e n t random sam ple gath ered , bus data have to be tested for norm ally and variability in o rd er th at h ypothesis te stin g reg ard in g th e ch ang es in the m eans of th ese data after the

im plem entation of th e sch em e can be perform ed. The F test, und er the sin g le-facto r analysis o f v arian ce (ANOVA) m odel, is a p relim in ary sta tistic to be used b e tw e e n the alternatives is

F *=MSTR/MSE where,

MSTR= treatment mean square

MSE=error mean square

And the appropriate decision rule to control the level of significance at EMBED Equation.α is:

If F* ≤ F(1-α ; r-1, nT-r), conclude Ho If F* ≥ F(1-α ; r-1, nT-r), conclude Ha where,

F (1-α ; r-1; nT-r) is th e ( 1-α ) 100 p e r c e n tile o f th e a p p r o p r ia te F distribution.

A detailed analysis was undertaken to determ ine if the F test leads to the conclusion that the factor level means μi differ. T hese d e ta ile d analysis include:

A. Estimation of a factor level mean .μi

The confidence limits for μi w ith confidence coefficient

1 - α are:

Y + t( 1-α / 2 ;n T -r)s{Y i}

w here

t ( 1 - α /2;nT - r)the tw o sided t-

distribution

s{Yi.}is th e e stim a ted stan d ard d e v ia tio n o f th e s a m p lin g distribution of Y (sample mean) B. Turkey method multiple compari­

sons of factor level means The T u rk ey m e th o d o f m u ltip le comparisons considers the set of all pairwise comparisons of factor level means which consists of estimates of all p airs D = μi - μi. T he family confidence coefficient of at least 1 -α are as follows:

W hen the distribution of the data depart from normality. The Kruskal- Wallis Rank Test can be used to test the w h ether the treatm ent means are equal. The test statistics of the Kruskal- Wallis Rank Test, X2KW, expressed as

With n i reasonably large (5 or more), X2kw is approxim ately a X2 random variable w ith r-1 degrees of freedom and having the following hypothesis test:

H o:μi=μ 2 =....μr Ha:not all μi are equal

The a p p ro p riate decision rule for controlling the risk of making a Type I error at α is:

If X2KW≤X2(1 -α ;r-1), conclude Ho IfX2KW>X2(l-α ;r-l), conclude Ha

3. METHODOLOGY

Since the prim ary objective of this study is to determ ine the effect of the MBSS on the operating characteristics, bus data w ere collected before and a fte r the im p le m e n ta tio n o f th e scheme.

T h e p rim a ry bu s data c o lle c tio n e m p lo y e d in th is s tu d y are th e following:

The boarding checks study. Most bus data w ere gathered using this data collection method. The data collector rode the bus being observed from the start of the route up to th e end of its journey and collected the necessary data along the way. Giannopoulos (1 9 8 9 ) d isc u sse d this m e th o d o f d e te r m in in g th e n u m b e r o f passengers inside the bus at various p o in ts o f th e r o u te a n d also to determ ine passenger boarding and alighting.

To m aintain the random ness of data, after identifying the bus service route to be surveyed at the beginning of the journey, the first bus to go was chosen as th e sa m p le . If p o s s ib le , th e surveyor should be the first passenger of the bus. Important data w ere listed down such as route serviced by the bus, time of start, seating capacity of the bus, and date of survey. As the bus stops to pick up and /o r drop off passengers, the exact moment the bus stops and the time it moves afterwards w ere noted down to determ ine the

tim e o f p a s s e n g e r b o a rd in g and alighting. Other reasons for stopping such as obstruction caused by other vehicles and intersection delay w ere also noted dow n. At the end of the journey, the surveyor made it a point to be the last or part of the last group of passengers to disembark. The exact tim e the journey ended was then noted down. The last step was to subtract bus seat capacity by one (tw o) as this was occupied by the surveyor (tw o surveyors), usually the seat near the door.

This study was perform ed during the w hole year of 1997 and up to June 1998. MBSS data w ere obtained after the scheme was implemented starting Novem ber 1 2 , 1997.

Bus speed and delay studies. Data needed for com putation of bus speed and delay w ere easily obtained by the surveyor by taking note of the time w hen the bus passes im portant road junctions or landmarks as well as the causes and duration of delays along the route. Distances w ere later obtained from a scaled map. O ther im portant o b s e r v a tio n s r e g a rd in g b us interaction with other vehicles on the road as well as ongoing activities and land use along the route w ere also listed down.

Bus volum e study. The p rim a ry purpose of this study is to determ ine the volume of buses passing through EDSA. The bus volume survey was conducted from 0630 to 0900 HRS in the morning and another one from 1630 to 1930 HRS in the evening.

T h re e bus volum e su rv ey s w e re

conducted during these periods: one in August 1997, Novem ber 1997, and in A p ril 1998. F ifte e n -m in u te intervals w ere used in determ ining the peak hour factor.

A fter processing the data collected, s e v e ra l b u s s e rv ic e o p e r a tin g characteristics w ere obtained:

A. Average travel speed. Average tra v e l s p e e d is e q u a l to th e distance of the segm ent being studied divided by the total travel tim e (in clu d es s to p tim e and running tim e) expended on the segment.

B. Average running speed. Average ru n n ing sp eed is equal to the distance of the segm ent being studied divided by the running tim e (o r m oving tim e) o f th e segment; and

C. P a s s e n g e r - k i l o m e t e r perform ance. All data regarding the com putation of passenger- kilom eter perform ance of buses were obtained w ith the surveyor on the bus. The com putation process was as follows:

(a) For the route segm ent of the study, every stop- and move­

time of the bus was known.

Since th e d ista n c e o f th e s e g m e n t w as g iv e n , th e average running speed could easily be com p u ted as the d is ta n c e o f th e s e g m e n t divided by total running time within the segment.

(b) K n o w in g th e n u m b e r o f passengers inside the bus at all tim es, th e p a sse n g e r-

kilometer performance of the bus every time the bus moves along the segm ent can be easily obtained as follows No. of Pax-Km = No. of passen­

gers inside the bus

x the average running speed x the running time

The average running speed for the whole segment was used since only the distance of the segm ent was know n and it was quite difficult to obtain the bus location every time the bus stopped and moved along the s e g m e n t for a d e ta ile d p a s s e n g e r - k i l o m e t e r computation.

Statistical techniques were then used to analyze the data as discussed in the previous section to arrive on a set of conclusions.

4. PUBLIC BUS TRANSPORT ALONG EDSA

Metro Manila’s PUBs. The public utility buses of M etro Manila are mostly imported. The buses are usually single- door types for air-conditioned unit and two-door types for non-air-conditioned units. The seating capacity of these buses range from 46 to 66, were the seats are usually arranged in fours or fives in a row, two-by-two or two-by- three facing the direction of motion.

T h e re are so m e m o d ific a tio n s, however, like the removal of some of the seats near the door to improve m ov em ent o f e m b a rk in g and disembarking passengers, or to provide

space for standing passengers. Bus fares are c o lle c te d by a c o n d u c to r / conductress inside the bus.

Bus routes. The major street network of Metro Manila consists mostly radial and circumferential routes. The radial routes radiate from the city of Manila toward other urban and suburban areas while circumferential routes traverse these radial routes. Urban centers that attract most daily urban trips are the Ayala and Ortigas business districts.

Figure 1 also shows the extent of bus route being studied w ith EDSA as the major segment of these route.

Urban bus service. Urban service in Metro Manila can be considered as local bus service since all stops along a route are being served. Along EDSA, buses stops on all bus stops and even on locations not designated as bus stops, although MBSS has modified this to some extent. With MBSS, the num bered buses can only stop in 10 of the 22 designated bus sto ps along EDSA.

However, as observed, buses were still able to stop at any point along the route e ith e r to d ro p off o r to p ick up passengers especially when there were no traffic enforcers around.

Major bus stops along EDSA are either situated on the near side or far side of an intersection or at locations w here there are major activity centers like shopping malls. Bus stops are also located near pedestrian overpasses for the convenience of the commuters. All bus stops are adjacent to the curb for direct and easy access by passengers.

Bus volume study. The ongoing Metro M anila U rban T ra n s p o rta tio n

12

Integration Study (MMUTIS 1996-1998) mentioned that the peak period in the morning occurs from 0700 to 1000 HRS, while the evening peak period occurs from 1600 to 1900 HRS. Three bus volume surveys were conducted during these periods: one in August 1997, November 1997, and in April 1998.

Table 1 shows the average results of the first two surveys while Table 2 shows the bus volume survey in April 1998 with bus numbering fully in place.

The peak hour volume has observed to o c c u r w ith in th e sp e c ifie d p eak periods. Not much different in bus traffic volume during the peak hour transpired after the implementation of the MBSS scheme. The data showed that

■ The p roportion of buses in the north-b ou nd and south-bound directions has a difference of no m o re th a n 10 p e r c e n t, w ith higher south-bound traffic during th e m o rn in g p e a k h o u r and higher north-bound traffic in the afternoon peak hour

■ The ratio of non-air-conditioned buses to air-conditioned buses is around 1:3

■ T he p r o p o rtio n o f n u m b e r 1 b u se s is q u ite lo w (9 to 10 percent) com pared to num bers 2 and 3 buses (40 to 48 percent) of the total buses flow during the estimated peak hour

■ W ith MBSS fully in place, th e volum e su rv ey in April 1998 showed that around 1 to 2 percent of buses travelling during the peak h o u r did n o t have n u m b ers;

roughly the same proportion had signs too inconspicuous to be seen.

T a b le 1. P u b lic b u s tr a n s it v o lu m e s u rv e y

Table 2 . P u b lic bus tra n s it volum e su rve y (A p ril 1 9 9 8 )

5. BUS OPERATING

CHARACTERISTICS FOR ANALYSIS OF THE MBSS EFFECT

The bus operating characteristics that w ere analyzed to m easure the MBSS effect include travel and stop times, average travel and running speeds, and average p a sse n g e r-k ilo m e te r performance.

Travel and stop times can provide facts about the effect of reduced num ber of stops on bus travel. Average travel and ru n n in g s p e e d s can p ro v id e information w hether improvements in bus travel have resulted, w hile av e ra g e p a s s e n g e r-k ilo m e te r performance can give insights into the revenue perform ance effect of the MBSS im p le m e n ta tio n on b u s operators. The study gave emphasis on Gil Puyat Extension to A urora Boulevard segm ent of EDSA. This segm ent is part of the routes being serviced by all buses passing EDSA regardless of their destination within Metro Manila.

Bus Travel Time and Delay. Table 3 shows the average travel (stop plus moving tim e) and stop time of air- conditioned buses before and after the implementation of the MBSS along the Gil P u yat E x te n s io n to A u ro ra Boulevard segment of EDSA. Figure 2 show s the scatterplots of the travel time versus stop time. The following observations can be derived from the presentations:

A. In general, the average travel has b e c a m e lo n g e r a fte r th e im plem entation of the MBSS by around 0.223 hr. (13.38 minutes) while average stop time increased by only aro u n d 0.07 h r (4.26 minutes).

B. The relatio nship b e tw e en the travel and the stop times of buses as represented by the trendlines, has s h ifte d u p w a rd a fte r th e implementation of the MBSS. The shift tells us that considering the same duration of time travel along the segment, travel time was equal to one hour before MBSS, while

stop time was around 0.444 hr. (27 minutes). After MBSS, it was only around 0.344 hr. (21 m inutes), or a savings of six m inutes on a one- h o u r to ta l tra v e l tim e. T his observation could be the result of Table 3. Average travel and stop times of air-conditioned buses

A v e ra g e S to p T im e H o u r ( m i n u t e s )

A v e ra g e T ra v e l T im e H o u r ( m i n u t e ) B e fo re M B S S 0 . 2 9 1 ( 1 7 . 4 6 ) 0 . 8 0 2 ( 4 8 . 1 2 )

A fte r M B S S 0 . 3 6 1 ( 2 1 . 6 6 ) 1 . 0 2 5 ( 6 1 . 5 0 )

Figure 2. Stop versus travel time of air-conditioned buses along a segment of EDSA

reducing the num ber of major bus stops a bus can stop along the segment under MBSS.

Since in totality th e average travel time and stop times have increased, MBSS was hardly a factor in improving air-conditioned bus m ovem ent along EDSA. However, the data showed that this schem e reduced average stop time by a few m inutes p er average total travel time.

ANOVA Model I was used to test the other operating characteristics. Using this m ethod, the before-and-after MBSS situations as w ell as the bus num bering system w ere considered as factor levels and have probability d is tr ib u tio n o f r e s p o n s e s ( th e in d e p e n d e n t v a ria b le s ). T h e se responses are assumed normal and of the same variance (standard deviation)

and o b s e r v a tio n s a re ran d o m ly obtained and are independent from observations of o th er factor levels.

H ence, to sta rt th e analysis, th e response variables such as the average travel and running speeds and the av e ra g e p a s s e n g e r-k ilo m e te r perfo rm an ce of buses w ere tested c o n c e r n in g th e ir p r o b a b ility distributions.

5.1 Test on the Distribution of Data

Table 4 shows the data regarding the average travel and running speeds and average p a sse n g e r-k ilo m e te r performance of the 60 air-conditioned buses, 30 before and 30 after the im p le m e n ta tio n o f th e MBSS, surveyed. There are several available tests to determine w hether the sample data came from a normal population.

One of these is the normal probability plot w herein the actual residuals of the variable are plo tted against its e x p e c te d re sid u a ls. If th e p lo t appears straight along the diagonal, the data can be considered as normal.

Figure 3 shows one of these plots using the residuals and e x p e c te d residuals of the average passenger- kilometer of the air-conditioned buses sampled after im plem entation of the MBSS. To strengthen this normality p r o o f f u r th e r , th e c o r r e la tio n betw een the actual residuals and the expected residuals w ere compared to a set of co rre la tio n values u n d e r normality as prepared by Looney and Gulledge, Jr. (1985). Table 5 shows th e s e c o r r e la tio n v a lu e s o f all v ariab les s tu d ie d w e re only th e average running speed (before MBSS)

and the average passenger-kilometer (after MBSS) of the buses passed the critical values. The average passenger- kilometer (before MBSS) was quite near the critical value and the rest of the variables w ere quite far off. Hence, the average passenger-kilometer for both cases (before and after) tend to behave normally.

Except for the average running speed (before MBSS), the other distribution of the average travel and ru n n in g speeds were not near normal. Hence, the average and running speeds was not further analyzed with regard to their variability because they seemed have failed one major criteria for normality.

T a b le 4 . A v e ra ge trave l and run n in g s p e e d s and a v e ra g e p a s s e n g e r - k i l o m e t e r p e r f o r m a n c e of a i r - c o n d i t i o n e d b u s e s

F ig u re 3 . P lo t o f re s id u a ls v s . e x p e c te d r e s id u a ls o f th e a v e ra g e p a s s e n g e r-k ilo m e te r p e rfo rm a n c e o f a ir-c o n d itio n e d b u s e s a fte r M B S S

T ab le 5 . T est fo r n o rm a lity u sin g th e c o e ffic ie n t of c o rre la tio n of a ir- c o n d itio n e d b u s e s o p e ra tin g c h a ra c te ris tic s .

T a b le 6 . F -T e s t fo r v a ria n c e s o f th e a v e ra g e p a s s e n g e r-k ilo m e te r p e rfo rm a n c e o f a ir-c o n d itio n e d b u s e s b e fo re a n d a fte r th e im p le m e n ­ ta tio n o f th e M B S S

T h e v a ria n c e s o f th e a v e ra g e passenger-kilometer were tested using the F-Test for equality of variance.

Table 6 shows F*>critical, therefore the variances of the average passenger-km perform ance of air-conditioned buses before and after the implementation of the MBSS somewhat differed. However,

th e value of F* is quite near th e Fcritical w hich tells us that the variability was not that large. As Figure 4 would show, their frequency polygons w ere almost similar in shape with only some shifts near the center.

F ig u re 4 F re q u e n c y p o ly g o n o f th e a v e ra g e p a s s e n g e r-k ilo m e te r p e rfo rm a n c e o f a ir-c o n d itio n e d b u s e s

Similarly, for non -air-co nditio n ed buses, the variables being studied are show n in Table 7 below. Again, the resu lts in Table 8 sh o w th a t the average passenger-kilometer garnered behave normally. Figure 5 presents the norm al probability plot of the actual residuals versus the expected residuals of the average passenger- k ilo m e te r o f n o n -air-c o n d itio n ed b u s e s s a m p le d b e fo re th e implem entation of the MBSS.

Table 7. The average tra ve l and ru n n in g speeds and average p a s­

s e n g e r-k ilo m e te r p e rfo rm a n c e o f n o n -a ir-c o n d itio n e d buses

T a b le 8 . T e s t fo r n o rm a lity u s in g th e c o e ffic ie n t o f c o rre la tio n o f n o n - a ir -c o n d itio n e d c h a ra c te ris tic s

F ig u r e 5 . N o r m a l p r o b a b ility p lo t o f e x p e c te d r e s id u a ls v s . a c tu a l r e s id u a ls o f th e a v e ra g e p a s s e n g e r - k ilo m e te r p e r f o r ­ m a n c e o f n o n - a ir - c o n d itio n e d b u s e s b e fo re M B S S

In Table 9 below, at a = .05, F* < F critical. Therefore th e variances of the ave. paxpkm perform ance of non-air- conditioned buses before and after the im plem entation of the MBSS did not differ.

As determined, the sample data of the average pax-kilometer perform ance w e re q u ite n o rm a l a n d e q u a l variability. W ith bus sam ples, the confidence interval range can now be obtained for the average pax-km a n d is sh o w n in F ig u re 6. T h e co nfid ence interval range for air- c o n d itio n e d b u se s s h if te d a b it dow nw ard after the im plem entation of th e MBSS while that for non-air- conditioned buses moved a bit upward T h e c o n fid e n c e ra n g e o f th e airconditioned buses is quite tight compared to non-air-conditioned buses because there w ere m ore samples taken (30 samples for air-conditioned buses, com pared to 14 for non-air - conditioned buses for both before and after the MBSS) of the former.

F ig u re 6 . E s tim a te d c o n fid e n c e in te rv a l (α =.0 5 ) o f th e a v e ra g e p a s s e n g e r -k ilo m e te r p e rfo rm a n c e o f a ir -c o n d itio n e d a n d n o n - a ir - c o n d itio n e d b u s e s b e fo r e a n d a f t e r th e im p le ­ m e n ta tio n o f th e M B S S

A statistical tests showed, the average travel and running speeds of buses did n o t b e h a v e n o rm ally, h e n c e w e cannot assume that the distribution of th e s e v a ria b le s a re n o rm a lly distributed. Just to give some insights into the effect of the MBSS on these variables, the sample means of the average travel and running speeds for all bus types are given in the Table

10.

T a b le 9 . F -T e s t f o r v a ria n c e s o f th e a v e ra g e p a s s e n g e r -k ilo m e te r f o r n o n - a ir -c o n d itio n e d b u s e s b e fo re a n d a fte r th e im p le ­ m e n ta tio n o f th e M B S S

Table 10. Average travel and running speeds of buses before and after the implementation of the MBSS

T h ro u g h th e av e ra g e tra v e l and ru n n in g sp e ed s w ere non-norm al th e r e a re o th e r n o n -p a ra m e tric s t a t i s t i c a l to o ls a v a ila b le to d e te rm in e the change in m eans of th e sam ple su ch as th e K ruskal - Wallis rank test. A pplying K ruskal W allis ra n k te s t to d e te r m in e d w h e th e r change in average travel and ru n n in g sp eed o ccu rred after th e application of the MBSS for both a ir - c o n d i ti o n e d a n d n o n -a ir- co n ditio ned buses, Table 11 gives th e result. As sho w n , th e average travel and ru n n in g speed s o f air- co nd itio ned buses decreased after th e ap p lication of th e MBSS w hile th a t fo r th e n o n -a ir-c o n d itio n e d buses rem ain th e same.

Table 11. Kruskal-Wallis Rank Test for changes in the average travel and running speeds

5.2 ANOVA MODEL I to Test the average Pax-Km Performance of the buses

Using ANOVA MODEL I, tw o tests w e re u n d e rta k e n on th e average passenger-kilometer perform ance of the buses. These are the following:

1. With two factor levels, (1) before MBSS, and (2) after MBSS, data on the average passenger-kilometers (the response variable) from the tw o studies were tested if changes occurred. This was done for both a ir-c o n d itio n e d a n d non-air- conditioned buses.

2. W ith buses under MBSS grouped into th re e , w e th in k of th ese groupings as factor levels. The test would like to determ ine how they have perform ed and w hether anyone among the three groups got w o rse/ b e tte r off from the schem e. This was done only for air-conditioned num bered buses.

Assuming that w e to control the risk of making Type I erro r at a= .05, Fcritical= 4 .0 0 7 S in ce F*<Fcritical w e conclude H0 that the m ens of the average passenger -kilometer of air conditioned buses along the segment m entioned did not differ after the im plem entation of the MBSS.

U sing th e sam e te s t on non-air- con d itio ned buses, F* = 1.323 < F critical = 4.2252, w e again conclude H0, that the means w ere equal or that the means of the average passenger- kilom eter p erfo rm an ce of non-air- conditioned buses before and after MBSS did not differ.

Table 12. F-Test fo r variances of th e average passenger-kilom eter perform ance of air-co n d ition e d buses

U nder th e MBSS schem e, the buses w e re d iv id e d in to th re e g ro u p s num b ered 1, 2, and 3. It w ould be interesting to know how each group has b e e n p e rf o r m in g u n d e r th e sch em e. H ow ever, it w as n o ted d o w n th a t th e a ssig n in g o f bus num bers to buses w ere not done proportionately. This was observed from the volum e survey in Table 3, w h ere th e N o .1 buses are few (9- 10%), com pared to Nos. 2 and 3 (40- 48%), of th e total buses operating d u r in g th e p e a k h o u r.

N evertheless, th e sam e nu m b er of sam ples, 10 from each group, was taken w hich in the previous section to ta le d to 30 sam ples. Table 13 below show s th e average passenger- km p e rfo rm a n c e of the n um bered buses un d er the MBSS. On the other h a n d , T a b le 14 s h o w s th e c o e ffic ie n t o f c o rre la tio n o f th e m eans of th e average passenger- kilom eter have passed the critical value show ing th at they ten d to be norm ally distributed. These can be fu rth e r bo lstered by the frequency histogram show n in Figure 7 w h ere th e average p a sse n g e r-k ilo m e te r p erfo rm an ce of the m ajority of the s a m p le s te n d to c lu s te r in th e middle.

Table 13. Average passenger-kilom eter perform ance of air-con ­ ditioned num bered buses under the M BSS

Table 14. A ir-conditioned num bered buses te st fo r norm ality

Figure 7. Sam ple frequency histogram of the average passenger- kilo m ete r of num bered buses

Table 15. Com parison of the average passenger-kilom eter variances betw een the num bered buses

Table 15 above shows the F-Test for variances of the num bered buses.

Only the variances of the buses 1 and 3 did not differ since the com parison passed the critical F-value. The other values being com pared have F -values a little over near the F critical though they were expected to pass the critical F -values if only m ore samples are taken.

It is n o w d e s ir e d to d e te r m in e w h eth er or n ot the m eans (pi) of the ave. pax-km of the num ber of equal.

The alternative conclusions considered w ere : H0: μ1 = μ2= μ3 and Ha : not all pi are equal Since F -0.4597

<Fcrit=3.4028, we concluded H0 that the means of the average pax-km of three bus types are equal. This results tells us that there was no relation existing betw een the numbering of the buses and t h e ir c o rre sp o n d in g average passenger-kilometer performance.

Table 16. The result o f the ANOVA fo r the three bus types

Since it has b een proven using th e F-test that no relation exist betw een the levels (the num bering of buses) and the dependent variable (the ave.

pax -k m p e r f o r m a n c e ) , d e ta ile d analysis of the factor level effects may not be necessary. However, additional analysis will be undertaken to show additional proof of w hat has just been proven. Further analysis include, (1)

Normal probability plot of estimated m eans of the average pax-km of the num bered buses, and (2) The Tukey method of multiple comparison of the ave. pax-km means of th e num bered buses (factor levels).

Normal Probability Plot of Estimated Factor Level Means. Since all sample sizes are equal (n=10), the normal p ro b a b ility p lo t can b e u s e d to determ ined factor level means. The ex p ected value o f th e ith ordered estim ate facto r level m eans, if all factor level m eans μi w ere equal, is a p p ro x im a te as (S ee N e te r, W a s s e rm a n ,a n d K u tn e r, 1 9 9 0 ) Expected value =Y + z[(i - .375)/ (r .25) X Sqrt (MSE/N), w here the square root term is the estim ate standard deviation of Yi., i= 1,2 and 3.

The estim ated factor level m eans Yi is p l o t t e d a g a in s t th e n o r m a l p ercentiles. A linear p a tte rn o f the p o in ts w as o b ta in e d s in c e th e e x p e c t e d v a lu e s a r e a lin e a r fu n ctio n of th e p e rc e n tile s. The e x p e c te d values w e re also p lo tted to serve as a reference line w h eth er h o w far th e e stim a te d m ean Yi dev iate from its e x p e c te d value.

The plots of th e data are sh o w n in Figure 8 below.

Figure 8. N orm al pro ba bility plo t of estim a ted m eans of passen­

ge r-kilom eter perform ance of the three bus types

The fact that the points of the means of the ave. pax-km performance of the three num bered buses w ere close to the reference line (expected value), these suggest that all means μ1 are reasonably equal In other words, result s u p p o rte d th e F-Test c o n d u c te d previously.

Multiple Comparison of factor level means (Turkey Method). To estimate all pairwise comparisons by means of the turkey procedure, a family confidence coefficient of 95 percent was used. From data, r = 3 and (n T - r) = 27 which could give us the required percentile of the s tu d e n tiz e d range d istrib u tio n q (.95;3,27)= 3.51. other necessary data were obtained as follows:

T = ( 1/ √2)(2.496) = 2.482 s2{D}=

1 9 0 60.66 (1 /9 + 1/9) = 2663.054, resulting to s {D }= 51.6048; hence, Ts{D}= 128.08

The p airw ise co n fid en ce interval w ith 95 p e rc e n t family co nfidence coeffient th ere fo re are:

-105.29=(340.49-317.70)-128.08≤ μ 2-μ1≤ (340.49-317.70+128.C8=150.87 -7 3 .1 0 = (3 7 2 .6 9 - 317.70) - 1 28.08 ≤ μ3-μ2 ≤ (372.69 - 317.70) +128.08 = 183.07

-9 5 .8 9 = (3 7 2 .6 9 -3 4 0 .4 9 ) - 128.08≤ μ3-μ1≤(3 7 2 .6 9 -1 2 8 .0 8 = 160.28

The pairw ise com parisons indicate t h a t th e m e a n s o f th e a v e ra g e passenger-kilometer perform ance of buses along the segm ent m entioned did not differ from each other. Since the confidence intervals cover the value zero, all of the differences are s ta tis tic a lly in s ig n ific a n t. No n u m b e r e d b u s ty p e has g a in e d significantly form the bus numbering a rra n g e m e n t. This resu lt against agrees w ith the earlier findings.

5.3 Kruskal-Wallis Rank Test:

fo r Average Travel and:

Running Speeds

Since it has b een sh o w n th at th e average travel and runn ing speeds d e p a r t fro m n o rm a lity , a nonparametric test called the Kruskal- Wallis Rank Test was em ployed to estim ate p ossib le ch an g es in th e means. The only assum ption in the population distribution being studied using the Kruskal-Wallis test is that

F ig u re 9 . A vera ge tra v e l spe ed of a ir-c o n d itio n e d n u m b e re d b u se s

F ig u re 1 0 . A ve ra g e ru n n in g s p e e d o f a ir-c o n d itio n e d n u m b e re d b u s s e s

they should be continuous and of the same shape (Neter, et. al., 1990). As Figure 9 and 10 show, the frequency histograms of the average travel and running speeds of the air-conditioned buses are quite similar.

Using the data of Table 17 it can be shown that the B. Kruskal-Wallis test X2 KW=3 .672 was lower than the X20.95,2)

= 5.991, hence we conclude H0 that the mean of the average running speeds of the num bered buses did not differ.

Doing the same process to the average

travel speeds of the num bered buses, X2KW=2.7587 was lower than th e X2 0.95,2) = 5.991.

Table 17. Average running speed of numbered buses and their Kruskal-Wallis ranks

These results show ed that non of the numbered buses got a better/worse deal in their perform ance since the hypo­

thesis that the means of their average travel and runnin g w ere the same cannot be disproved.

6. CONCLUSION AND RECOMMENDATIONS

The study aimed to determine the effects of MBSS on bus operating performance along EDSA. Using the bus operating characteristics such as travel and stop tim es, average travel and ru n n in g s p e e d s, an d averag e passen g er- kilometer perform ance of buses along the Gil Puyat Extension to Aurora Boulevard se g m e n t of EDSA, th e following conclusions were derived from the study:

A. In general, the mean of travel and stop tim es increased after the

im p lem e n ta tio n o f th e MBSS, although there was a noted decrease in stop time for a given value of travel time may be attributed to the decrease in the number of stops the bus are allowed to stop;

B. One possible reason why average travel and running speeds do not behave normally is that due to MRT and other construction activities, closing of a d d itio n a l lanes so m etim es o c c u r in an unpredictable m anner thereby causing bottlenecks and reducing average travel and running speeds.

This is not true with respect to the average p a sse n g e r-k ilo m e te r performance of the buses, which behaves normally, since daily bus p a sse n g e rs rem ain th e sam e w hether there is congestion or not along EDSA.

C. Using analysis of variance with a confidence level of α .05, the average p a sse n g e r-k ilo m e te r gained by both the air-conditioned and non-air-conditioned buses did not differ after the implementation of the MBSS. However, th is may not be true of the average passenger- kilometer performance of the buses of the entire route they service or on daily basis;

D. Using analysis of variance with a confidence level α = .05, for the air-conditioned num bered buses, no group got better/w orse off in terms of the average passenger- kilometer performance. However, as the volume survey would show, there were quite a small percentage of num ber 1 buses in operation.

Since one of the primary purposes of the scheme is the decongest bus

stops, equal assignment of buses to each group should be done. In the present situation, if equal proportion of buses is considered, number 1 buses will be disadvantaged. Hence, a reassignment of bus stops to the three bus groups should be looked into before equal proportioning of buses is done.

E Using Kruskal-Wallis Rank test, the average travel and running speeds of air-conditioned buses decreased after the implementation of the MBSS while that for the non-airconditioned buses did not differ. Also, for the air- conditioned numbered buses, no differences among the groups were noted in terms of their average travel and running speeds.

This study only considered the effect of the MBSS on bus operating characteristics.

However, there are other stakeholders involved whose interests are as important as the buses and bus operators, such as the bus riding public and other vehicles using EDSA. With buses limited to fewer stops than before, waiting time for bus commuters certainly became longer. It would also be interesting to know how traffic flow on the point of view of private vehicles has changed after the MBSS implementation.

ACKNOWLEDGEMENT

The author would like to acknowledge the support provided by the University Research Coordination Office of DLSU for providing the necessary funds for the research on air-conditioned buses. The author would also like to acknowledge Mr. Law rence Alcon, Mr. Rommel Icasiano, and Mr. Virgilio S. Vinzon II, all DLSU civil engineering students, for

sharing with the author their data on non-air-conditioned buses.

REFERENCES

D epartm ent of T ransportation and Communication (DOTC) Metro Manila Urban Transportation Integration Study (MMUTIS), in c o o p e ra tio n w ith Ja p a n In te rn a tio n a l C o o p e ra tio n Agency (JICA), Progress Report

1, January 1997.

Fillone, A. M. and Fukuda, A. (1989) Service Perform ance of City Buses in M etro Manila, in:

Proceeding of th e Second Regional Sym posium on Infrastructure Planning in Civil Engineering, February 20-21, Manila, Philippines.

Giannapoulos, G. A. (1989) Bus Planning and Operation in Urban Areas: A practical Guide, Avebury, Singapore.

Jordan, A. C. and Turnquist, M. A. (1979), Zone Scheduling of Bus Routes to Improve Service Reliability.

Transportation Science, Volume 13, No. 3, August 1979.

Looney, S. W. and Gulledge Jr., T. R., (1985), Use of the Correlation C oefficient w ith Norm al Probability Plots, in: The American Statistician 39, pp. 75- 79.

Neter, J., et. al., (1990), Applied Linear Statistical Models, 3rd ed., IRWIN Tanaboriboon, Y. (1992), An overview

and future direction of transport demand management in Asian m etropolises, in: Regional Development Dialogue, Vol. 3, No. 3, Autumn Issue.

23