The Relationship Between Transit Asset Condition and Service Quality (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

C-1 Case Studies Case Study Approach The research team conducted four case studies to demonstrate the asset condition/service qual- ity framework described previously. The case studies were selected in consultation with the project panel and were intended to encompass as wide a set of operating characteristics and assets as pos- sible. In the summaries that follow, the case study transit agencies remain anonymous, but are represented geographically as follows: • New England Transit Agency Heavy Rail Line • Western Transit Agency Bus Route • Mid-Atlantic Transit Agency Heavy Rail Line • New Zealand Bus Systems To conduct each case study, the research team first met with the case study transit agency to confirm the scope of the case study, review the condition/service quality framework described in Chapter 3, and collect data. The research team then prepared a brief memorandum confirming the scope of the case study. Research team members then performed the analysis for the case study. The case study results were described in write-ups detailing case study scope, inputs, outputs, and out- comes. Draft write-ups were provided to the participants and revised based on participant concepts. The following sections summarize each of the case studies, as well as overall findings. Case Study 1: New England Transit Agency Heavy Rail Line Context For the New England Transit Agency case study, the research team focused on computing the EJT for a single rail line. Specifically, EJT was used to compare the level of service on the rail line before and after two significant improvements—new vehicles and track upgrades—were made to the line in the late 2000s. The year 2006 is used in this analysis to represent the level of service before the fleet replacement as this was the last year before new cars were introduced. By 2012 all the old cars (built between 1978 and 1980) had been replaced and phased out, so this year is representative of service levels following the replacement project. Also during this time, signifi- cant upgrades were made to the track. Using the EJT model, this case study shows the effects of vehicle and track improvements on effective passenger journey time and cost. Analysis Approach As stated above, the goal of this case study was to use the EJT model to show the effects of vehicle and track improvements on effective passenger journey time and cost. The main portion A P P E N D I X C

C-2 The Relationship Between Transit Asset Condition and Service Quality of the effort in this case study was using data provided by the New England Transit Agency, and found in other available sources, to compute each of the parameters required for the EJT model. The headway standard deviation was approximated based on the transit agency’s pub- lished schedule and on-time performance data, assuming headways and run times are distrib- uted based on a gamma distribution, consistent with the approach used by Richter, Ilzig, and Rudnicki (2009) and Ruan and Lin (2009). Once these parameters were determined, the research team introduced variations in the parameters to compare level of service in 2006 and 2012. The three sources of variation were as follows: • Vehicle failure rate. The failure rate was slightly lower in 2012 due to the new vehicles that replaced the old fleet. • In-vehicle comfort factor. A lower value was used in 2012 than 2006, representing customer perception of the old vs. new vehicles. As discussed in Chapter 3, a wide range of values can be found in the literature. Here an adjustment factor of 1.02 was used for 2012 (old vehicles), given the operating characteristics for the line being modeled appeared most similar to those described by Wardman and Whelan (2001), in which a value of 1.02 was calculated. For this reason, and to avoid overstating the effect of customer perceptions, a low value was used for the in-vehicle comfort factor. • Travel times. The travel times were longer in 2006 due to speed restrictions in place because of track issues. Results Table C-1 summarizes the EJT results incorporating both the vehicle and track improve- ments. This table shows the minutes per passenger and minutes per passenger mile for peak and off-peak travel times. In addition, the table shows the annual benefit of the improvements to vehicles and track. This value is calculated assuming a value of time of $12.55/hr, the FHWA personal value of time in 2014 dollars. As indicated in the table, the improvements to track and vehicles are predicted to save approximately 1.7 minutes per passenger mile in peak periods and 1.2 minutes per passenger mile in off-peak periods. The total annual benefit is predicted to be approximately $18 million dollars. This benefit is not actual dollars earned or saved, but only used as an expression to standardize the value of time. Table C-2 summarizes the results in the scenario that vehicles are improved, but track is left unimproved. For this scenario, the failure rate and In-Vehicle Comfort Factor inputs reflect the old and new vehicles in the respective years. However, the travel times used to show track Description Peak Off Peak Minutes Per Passenger Minutes Per Passenger Mile Minutes Per Passenger Minutes Per Passenger Mile 2006 21.031 8.36 27.562 10.95 2012 16.630 6.61 23.359 9.28 Annual Benefit $ 9,221,690 $ 8,806,808 Total Annual Benefit $ 18,028,497 Table C-1. Summary results—improvement in vehicles and track.

Case Studies C-3 improvements remain the same. As shown in the table, the annual benefit for this case is approx- imately $1 million. The two scenarios illustrated above demonstrate there is a significant benefit in journey time from changes to the line made between 2006 and 2012. An annual benefit of $18 million is predicted for the line as a result of improving track conditions and replacing the vehicle fleet. A large portion of the benefit can be attributed to the reduction in slow zones resulting from track improvements. Replacing the vehicle fleet has also resulted in gains. The new cars are assumed to enhance the customer experience, reducing effective travel time cost slightly. Also, the new cars are less likely to fail in service. However, the lower failure rate per car is partially offset by the greater number of cars per train. As a result, the benefit of replacing the fleet, though positive, is much smaller than that from improving track conditions. Case Study 2: Western Transit Agency Bus Route Context For the Western Transit Agency case study, the research team focused on a single bus route. The bus route was selected for its long length and the regularity of the schedule—buses stop at each of the major stops throughout the day with only a few exceptions. While there are numer- ous minor stops along the route, boardings and alightings were aggregated based on 15 major stops in the model. Analysis Approach The case study analysis included the following steps: • Determine parameters for vehicle running time, running time distribution, and vehicle fail- ure rate. The research team used the route schedule and information from the transit agency on technical failures and total mileage across the system. • Test the EJT model for a representative bus route on the Western Agency network. Using data provided by the transit agency on the failure rate, failure duration, and number of people on a bus, the team calculated a baseline journey time on the representative route. Then, to see how the journey time might change in the future, the team simulated the effect on EJT of potential, future deterioration in condition. • Compare the effects of changing asset condition predicted using the EJT model to the more approximate calculation of user costs in the TAPT model. Substitute EJT model results into the TAPT model to evaluate the effect of a more refined calculation of service quality impacts. Description Peak Off Peak Minutes Per Passenger Minutes Per Passenger Mile Minutes Per Passenger Minutes Per Passenger Mile 2006 21.031 8.36 27.562 10.95 2012 20.797 8.26 27.329 10.86 Annual Benefit $ 490,315 $ 488,219 Total Annual Benefit $ 978,534 Table C-2. Summary results—improvement in vehicles only.

C-4 The Relationship Between Transit Asset Condition and Service Quality Results EJT Model Parameters for vehicle running time, running time distribution, and vehicle failures were used in the spreadsheet model to compute EJT in terms of minutes and dollars per passenger, given current conditions and under different scenarios in which buses are allowed to deteriorate over time. Table C-3 summarizes the journey time results. The table shows journey time including and excluding adjustment for customer perceptions for the existing route; and for increases of 5, 10, and 15 years in fleet age. As noted previously, currently the fleet is relatively new. The table shows that EJT is predicted to be 72.47 minutes at present. Journey time per passen- ger is expected to increase to 78.23 minutes if the fleet is allowed to age 15 years, an increase of approximately 6%. This increase is largely a result of applying the in-vehicle comfort adjustment factor of 1.2 (which applies to IVT only and not wait time). If the adjustment factor is omitted, the increase is approximately 0.25%. Figure C-1 translates the results from Table C-3 into a cost in terms of dollars per passenger based on a personal value of time of $12.55 (FHWA default value for 2014). As shown in the figure, an increase of 5 years in fleet age increases journey cost by just a penny per passenger. Increased Vehicle Age Relative to Current (years) EJT (Minutes Per Passenger) Excluding Customer Perceptions Including Customer Perceptions 0 72.47 72.47 5 72.53 72.53 10 72.59 75.38 15 72.65 78.23 Table C-3. EJT predictions for Western Transit Agency bus route. 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 In cr ea se d Co st P er P as se ng er ($ ) Increased Age Relative to Current (years) w/o In-Vehicle Comfort Factor With In-Vehicle Comfort Factor 0 2 4 6 8 10 12 14 16 Figure C-1. Increased cost per passenger due to aging buses.

Case Studies C-5 Further deterioration results in more significant increases, up to $1.20 per passenger for a 15-year increase in age accounting for the adjustment factor. Comparison to TAPT Although the per passenger cost increase resulting from asset deterioration may seem modest, it is a greater cost than that predicted by TAPT, given the simplified user cost model it uses. As a final step in the analysis, the research team converted the per passenger cost into a cost per bus and sub- stituted this cost into TAPT, replacing its cruder estimate of user costs. Figure C-2 shows the lifecycle cost predicted for a Western Agency 40-foot transit bus with the previous model (solid line) and the EJT results substituted in (dotted line). With the EJT results included, user costs increase signifi- cantly as the fleet ages, and the cost-minimizing replacement age decreases from 14 years to 9 years. Although it is clear that the user costs predicted from the EJT model are greater than those predicted with the TAPT model, several parameters are incorporated in the prediction of total lifecycle costs and it is not necessarily the case that adding EJT model results will always result in a shift in the cost-minimizing replacement age as shown here. Further, in this particular case the TAPT results obtained previously for the Western Transit Agency from Robert et al. (2014b) were not recalibrated to account for changes in fleet composition since the prior case study. Case Study 3: Mid-Atlantic Transit Agency Heavy Rail Line Analysis Approach The Mid-Atlantic Transit Agency has a comparatively rich set of data resources that the research team leveraged in performing the case study, including farecard data, a detailed incident log, on-time performance data, and quarterly customer satisfaction survey results. Of particular Cost-Minimizing Age Figure C-2. Lifecycle cost curves.

C-6 The Relationship Between Transit Asset Condition and Service Quality note is the transit agency’s farecard data. Through its farecard system, the transit agency recently began obtaining detailed travel time data for its rail system, with records of the origin, destination, start time (time of entry through the fare gate), and end time (time of exit through the fare gate) of each trip. This data supports calculating a journey time directly and thus allows for a more or less direct comparison of modeled and actual journey time results. The research team obtained journey time data for two specific dates: a representa- tive “good” day with few delays and a representative “bad” day with multiple guideway and vehicle failures. The “bad” day selected for analysis was one in which multiple guideway failures, related to declining asset condition, caused significant delay. Figure C-3 shows a map of the North Line. The name and route geometry of this line have been changed to keep the case study transit agency anonymous. The map shows the station numbers, transfer points, and the incidents that occurred on the “bad” day. Specifically, at the beginning of the day, a track problem (an arcing insulator and power cable) was encountered at Station 6 on the North Line, resulting in single-track service in the vicinity of this station (between Stations 5 and 8). Subsequently, a similar problem was found several stations away at Station 12, resulting in single-track service there, as well (between Stations 10 and 13). Thus, for the morning peak period, the North Line operated on a single track on much of the northern portion of the line. Given the constrained capacity of the line, some trains were turned south of the single-track portion of the line, further reducing capacity and increasing delay. Customers on the North Line were delayed 20–45 minutes on average throughout the morning commute, based on the incident log and local news outlets. In addition to this incident, the system experienced 12 other vehicle-related issues on the same day. Based on the initial discussions with the Mid-Atlantic staff and the data available for analysis, the research team performed the following analyses as part of the case study: • North Line EJT Model. Developed an EJT model for the North Line representing typical operations of the line, as well as operations during the incident on the “bad” day. Figure C-3. Mid-Atlantic Agency North Line map.

Case Studies C-7 • Analysis of Farecard Data. Estimated average North Line journey time using data for the “good” day. Calculated effects on journey time for each incident on the “bad” day, comparing journey times encountered during each incident to the same period of the “good” day. • Comparison of Modeled and Actual Results. Compared EJT model results to farecard data to assess differences between modeled results and actual journey times recorded by the farecard system. Results North Line EJT Model The spreadsheet model was used here to estimate EJT for the Mid-Atlantic Transit Agency’s North Line. The following are key parameters used by the model and a description of how they were determined. • One requirement for the model is an Origin-Destination (O-D) matrix. Previously this was synthesized from data on boardings and alightings by station. Here the O-D matrix was deter- mined using the farecard data for the “good” day, substituting in the appropriate North Line transfer points for the origin or destination of a trip that used multiple lines. • Train headways and run times between stations were obtained from the Mid-Atlantic Transit Agency. The standard deviations of headways and run times between stations were estimated based on the Mid-Atlantic Transit Agency’s published schedule and on-time performance data, assuming headways and run times are distributed based on a gamma distribution, con- sistent with the approach used by Richter, Ilzig, and Rudnicki (2009) and Ruan and Lin (2009). • Failure rates and durations for guideway and vehicles were estimated by analyzing incidents for the “bad” day and supported by supplemental analysis of incident effects from the fare- card data for that day described in the next section. During the month when the “bad” day occurred there were 151 vehicle failures resulting from mechanical causes on the North Line and 20 guideway failures. Vehicle failures were estimated to cause 10 minutes of delay per pas- senger on average and to affect two additional trains for each incident (with secondary delays of 5 minutes per passenger). Guideway delays were estimated to last 17.6 minutes on average. • Train capacity was determined based on data reported by the Mid-Atlantic Transit Agency to the NTD. • In order to maintain consistency with the New England Agency case study, also a rail case study, the value used for the in-vehicle comfort factor on deteriorated vehicles was 1.02. Table C-4 summarizes the EJT results by trip component and period of the day. Component of Time Period AM Peak Midday PM Peak Off Peak Average Unadjusted Buffer 3.9 5.7 3.6 5.8 4.6 Wait 1.9 6.8 1.9 6.8 4.4 In-Vehicle 17.1 12.8 14.3 13.9 14.0 Total 22.9 25.3 19.8 26.5 23.0 Adjusted Buffer 5.0 7.4 4.7 7.5 5.9 Wait 3.6 12.9 3.7 12.9 8.4 In-Vehicle 17.3 15.0 14.3 13.9 14.2 Total 25.9 35.3 22.7 34.4 28.5 Table C-4. EJT per North Line passenger by time of day (shown in minutes).

C-8 The Relationship Between Transit Asset Condition and Service Quality As summarized in the table, the EJT for the North Line averages 28.5 minutes incorporating various adjustments for customer perceptions. Of this total, approximately half is for time spent in vehicles, 30% is for wait time, and the remainder is buffer time. The greatest EJT is calculated for the midday period when train frequencies drop. Effects of changes in asset condition can be modeled in two basic ways. One approach is to change the failure rate for vehicles and/or guideway to calculate how change in average failure rates affect the average. A second approach is to approximate the effect of an event that might occur if an asset fails, knowing that such failures may become increasingly likely if system condi- tions deteriorate. Table C-5 shows results for different scenarios for the AM peak reflecting these two approaches. The base case is the same as that presented in Table C-4. In the scenario labeled “Failures Halved,” the rate of vehicle and guideway failures has been reduced to half of the default; in the case of “Failures Doubled,” it has been doubled. For the final scenario labeled “Guideway Incident,” the effect of the “bad” day incident was simulated by double train headways and cutting the number of trains by half. In addition, the run time was increased by 40 minutes in each direction in the area affected by the “bad” day incident, based on information published in Mid-Atlantic Transit Agency’s Daily Service Report for the incident and from news reports on that day. The table shows that large changes in the failure rate result in only modest changes in EJT (approximately a 1% change for halving or doubling the rate). However, in the case of the Guideway Incident scenario, reducing capacity and increasing train running time has a signifi- cant increase on EJT, approximately doubling the adjusted total. Analysis of Farecard Data The farecard data for the “good” day and the “bad” day was analyzed to determine journey times for the North Line and to assess the effect of guideway and vehicle events, particularly the guideway incident on the “bad” day. The data includes the origin, destination, start time, and end time of each trip on the rail system. One challenge in using the data to model a single line is that many trips involve multiple lines. To capture the portion of each trip made on the North Line, the effective North Line origin and destination were determined (with one or the other being a transfer point), and the North Line portion of the journey was estimated by using the average value for other pas- sengers with that actual origin and destination during the half-hour period the trip was made. Component Scenario Base Failures Halved Failures Doubled Guideway Incident Unadjusted Buffer 3.9 3.8 4.0 4.1 Wait 1.9 1.9 2.0 3.2 In-Vehicle 17.1 17.0 17.0 30.5 Total 22.9 22.7 23.0 37.8 Adjusted Buffer 5.0 5.0 5.1 5.4 Wait 3.6 3.6 3.8 6.1 In-Vehicle 17.3 17.3 17.3 38.8 Total 25.9 25.9 26.2 50.3 Table C-5. EJT per North Line passenger, AM peak for selected scenarios (shown in minutes).

Case Studies C-9 For instance, for a trip from Station 12 on the North Line to Station 15 on the East Line made beginning at 8:05 AM, one would most likely transfer between lines at Station 16 on the North Line. For this trip, the North Line portion of the trip would be Station 13 to Station 16, and the time for this portion of the trip would be estimated based on the travel times for other passengers who made a trip from Station 13 to Station 16 (without making a transfer) between 8:00 AM and 8:30 AM. Using this approach, the research team compiled the actual journey time by period of the day for the “good” day and the “bad” day. Tables C-6 through C-8 show results for the AM peak period (5:00 AM to 9:00 AM), midday period (9:00 AM to 3:00 PM), PM peak period (3:00 PM to 7:00 PM), and other times. Also, the tables show results for the entire day and for the period when the guideway incident occurred on the “bad” day, from 5:00 AM to 3:30 PM. Table C-6 shows passengers by period. Table C-7 shows total journey time. Table C-8 shows the estimated portion of the journey time spent on the North Line. Tabled C-6 shows that the North Line carries approximately 284,000 trips on a typical day. On the “bad” day, trips were slightly greater than the “good” day during the AM peak, but down significantly during the midday period and down approximately 3% for the day. This suggests that some passengers may have canceled or shifted their trips in response to the guideway inci- dent on the “bad” day, although the extent to which this may have been the case is difficult to verify, given day-to-day fluctuations in ridership. Table C-7 shows that trips involving the North Line are approximately 26 minutes long on average, including the North Line portion and time spent on other lines. As shown in Table C-8, for a typical day, the average journey time specifically on the North Line averages approximately 19 minutes both during the AM peak and overall. On the “bad” day, the average journey time during the AM peak was nearly 35 minutes, an increase of 84% over the average. The guideway incident that occurred that day persisted until approximately 3:30 PM. Over the entire period Period “Good” Day “Bad” Day Difference Percent Change AM Peak 74,907 80,245 5,338 7% Midday 74,282 57,499 -16,783 -23% PM Peak 99,612 105,285 5,673 6% Other Times 35,273 30,703 -4,570 -13% During Guideway Incident 157,787 144,537 -13,250 -8% Total 284,074 274,349 -9,725 -3% Table C-6. North Line passengers by period of day, “good” day and “bad” day. Period “Good” Day “Bad” Day Difference Percent Change AM Peak 26.4 43.5 17.1 65% Midday 25.4 35.2 9.8 39% PM Peak 25.0 29.5 4.5 18% Other Times 27.8 34.7 6.9 25% During Guideway Incident 25.9 39.8 13.9 53% Total 25.9 35.4 9.5 37% Table C-7. Average journey time for North Line passengers by period of day, “good” day and “bad” day (shown in minutes).

C-10 The Relationship Between Transit Asset Condition and Service Quality from 5:00 AM to 3:30 PM, the North Line journey time averaged approximately 31 minutes— 69% greater than average. Figure C-4 is a heat map showing the data for the AM peak on the “good” day and using dif- ferent colors to represent the North Line journey time between each North Line origin (hori- zontal axis) and destination. Figure C-5 shows the same chart for the AM peak on the “bad” day, illustrating the greatly increased times experienced by passengers journeying on the Northern portion of the line (between Stations 13 and 1). Figures C-4 and C-5 show that the event on the “bad” day caused significant delay, but the effects are averaged out to some extent, because not all passengers were affected equally. Origins D es tin ati on s Figure C-4. North Line travel times, AM peak – “good” day. Period “Good” Day “Bad” Day Difference Percent Change AM Peak 19.0 34.9 15.9 84% Midday 18.1 27.2 9.2 51% PM Peak 18.1 21.9 3.7 20% Other Times 20.7 26.5 5.8 28% During Guideway Incident 18.6 31.4 12.9 69% Total 18.7 27.3 8.7 46% Table C-8. Estimated average journey time for North Line passengers by period of day, “good” day and “bad” day—north line portion only (shown in minutes).

Case Studies C-11 Figure C-6 shows the difference between the 2 days. Cases where there was little or no increase are shown in green. Increases of 5–10 minutes are shown in yellow, 10–20 minutes in orange, 20–30 minutes in red, 30–40 minutes in purple, and over 40 minutes in black. As the figure illustrates, passengers boarding at Station 7 and headed inbound (just on the outbound of the start of single-track service) experienced the greatest delays during the guideway incident on the “bad” day. Comparison of Modeled and Actual Results Two basic questions were posed in comparing modeled and actual results for the Mid-Atlantic Transit Agency’s North Line: • How well does the EJT model predict actual performance of the North Line on a typical day; and • How well does the model predict the level of service impacts of the guideway incident on the “bad” day? Regarding the predictions of the EJT model for a typical day, overall the model appears to provide a very good approximation of the performance of the North Line on a typical day. As detailed in Table C-4, the EJT model predicts an observed journey time (unadjusted wait time plus IVT) of 19.0 minutes per passenger for the AM peak period and 18.4 minutes overall for the day. Table C-8 shows that the estimated North Line journey time for the “good” day was Origins D es tin ati on s Figure C-5. North Line travel times, AM peak – “bad” day.

C-12 The Relationship Between Transit Asset Condition and Service Quality 19.0 minutes for the AM peak period and 18.7 minutes overall for the day. Only the North Line portion of the journey time was predicted and that portion of the trip was estimated in the case of the farecard data. Also, the farecard data does not detail wait time versus IVT, does not adjust time for customer perceptions, and does not provide information on buffer time passengers need to allow (although this could certainly be estimated by analyzing multiple days’ worth of farecard data). The EJT model also appears to do a reasonable job of approximating the effect of the guide- way incident on the “bad” day. The Mid-Atlantic Transit Agency reported that portions of the North Line were single-tracked during the incident and that passengers incurred delays of up to 40 minutes during the AM peak period. The 40-minute estimate is higher than the average delay of approximately 30 minutes calculated from the data and shown in Table C-6, but appears representative of the worst case scenario as depicted in Figure C-6. To simulate the incident in the EJT model, the headways on the line were doubled and 40 minutes of running time were added in each direction in the area that was single-tracked. The resulting model predicts an unadjusted journey time of 33.7 minutes, as shown in Table C-5. Based on the farecard data, calculations show that the actual journey time (North Line portion only) was 34.9 minutes. Overall, the model matches the data relatively well, but there are several caveats. Because the delay was added to train running times, most of the delay was incurred as IVT. Accounts Origins D es tin ati on s Figure C-6. Difference in AM peak North Line travel times, “bad” day vs. “good” day.

Case Studies C-13 from the “bad” day suggest that most passengers were instead delayed in the station while waiting to board the train. The discrepancy in delay location does not affect the total signifi- cantly, but given the varying adjustments for waiting and IVT, the EJT model likely under- states the effective time incurred. Further, as shown in Tables C-7 and C-8, the incident resulted in a greater delay to overall journey time than to the North Line portion alone. To the extent that the EJT model predicts time for a single line it thus understates incident effects in this regard, as well. Case Study 4: New Zealand Bus Systems Context New Zealand offered a unique opportunity to analyze service quality of bus services provided by private-sector operators. The Transport Services Licensing Act of 1989 gave passenger trans- port operators the right to register and operate commercial services. Services that can make a profit from farebox recovery are therefore provided on a fully commercial basis, without finan- cial assistance or subsidy. These tend to be longer distance scheduled services, such as inter- city routes. Higher frequency metropolitan services are not profitable and are partially public funded. These services are tendered by regional councils (except in the case of Auckland Trans- port), with funding being provided both by regional council rates (a form of property-based taxes) and from the National Land Transport Fund to regional councils. Because of these commercial tensions, New Zealand has found it necessary to require mini- mum vehicle standards and monitor the delivery of service quality. Prior to the 1990s, metro- politan services were provided by publicly accountable entities. With the legislative change to the commercial model, much of the bus fleet in at least one city was replaced “overnight” by older buses, which patrons complained were noisy and uncomfortable and belched diesel fumes. For several years, condition and service quality declined along with patronage. The case study considers bus services across New Zealand. However, smaller provincial areas are excluded from the analysis. For example, Christchurch is a subset of the Canterbury region and, therefore, bus fleets not servicing Christchurch are not included. Most bus trips in New Zealand are made in Auckland, Wellington, and Christchurch. In 2009/10, out of a total of 125.6 million trips made on public transport, 101.2 million trips were made on buses. Eighty-eight percent of all bus boardings were made in New Zealand’s three largest public transport markets: Auckland (47%), Wellington (24%), and Canterbury (17%). This case study also presented the opportunity to explore an alternative (or enhancement) to the EJT model. NZTA’s EEM (2016) describes a range of user benefits and provides a mecha- nism for calculating the economic benefits of making improvements to transit services, such as reliability improvements, increased service frequency, interchange reduction, and other user benefits such as comfort. Other parameters include bus features and infrastructure provision relating to bus stops/shelters and bus stations. Of interest for this case study is whether the parameters and associated calculations from the EEM can be used as an alternative to the user benefit calculations that can be done with the EJT model, or whether information from the EEM can be used to enhance the parameters used as input in the EJT model. User benefit factors in the EEM are usually based on a willingness-to-pay value derived from stated preference surveys or on values derived for similar service improvements in other areas. Those attributes in the EEM relating to buses are listed in Table C-9. It is observed that although these factors are not linked to a state of good repair model, they do influence operational activ- ity and associated costs, such as regular cleaning or investment in additional on-bus facilities or equipment and could therefore be used in a lifecycle cost analysis.

C-14 The Relationship Between Transit Asset Condition and Service Quality Analysis Approach The case study analysis focused on the following two main aspects: • Explore the EEM as an enhancement or an alternative approach to EJT. To determine if user benefit factors from the EEM could be used to verify or enhance perfection factors in the EJT model, the EEM was used to compute an IVT for comparison with the In-Vehicle comfort factor used in the EJT model. To determine if the EEM could be an alternative approach to the EJT model, a benefit dollar value was computed from the EEM and compared to results from the EJT model as well. • Develop adjustment factors to address customer perceptions of asset conditions based on time series analysis. Using customer satisfaction data, the following relationships were examined: – Analysis of customer satisfaction and average fleet age – Bus quality vs. age. Results EEM Analysis The annual user benefit of operating a bus in good condition relative to one in poor condi- tion was calculated using a hypothetical example of a bus which is old and poorly maintained compared to one which is in good condition. This calculation incorporated the valuation, in in-vehicle time (IVT) minutes, of various attributes (see Table C-9) as well as the New Zealand standard travel time values for bus users. Table C-10 presents the annual user benefit results. Adjustment Factors to Address Customer Perceptions: Time Series Analysis Figure C-7 presents analysis of the customer satisfaction trends against average fleet age data for 2016. This analysis sought to determine whether there is any discernible relationship between average fleet age and overall customer satisfaction or other satisfaction measures such as reli- ability (given that reliability is assumed to be related to fleet age and condition). Attribute Sub-attribute Valuation (IVT minutes) Comment Cleanliness Litter 0.4 mins No litter compared with lots of litter Windows 0.3 mins Clean windows with no etchings compared with dirty windows and etchings Graffiti 0.2 mins No graffiti compared with many instances of graffiti Exterior 0.3 mins Very clean everywhere compared with some very dirty areas Interior 0.3 mins Very clean everywhere compared with some very dirty areas Comfort Legroom 0.2 mins Space for small luggage compared with restricted legroom and no space for small luggage Ventilation 0.1 mins Push-opening windows giving more ventilation compared with slide opening windows giving less ventilation 1.0 mins Air conditioning Table C-9. User benefits from EEM and their valuation.

Case Studies C-15 If there was a real relationship between the relative age of the fleet and reliability as experi- enced by customers, one would expect satisfaction to decline as bus fleets get older. However, in this high-level view, the opposite is the case, indicating that other factors are at play in the satisfaction data, such as • Traffic conditions—while Auckland has the youngest fleet, it also has the most congested traffic conditions • Driver behavior or other intangible factors relating to the journey • How well (or poorly) buses are being maintained in each fleet • General differences in perspectives among the three cities In addition, monthly survey data was correlated with the age of individual vehicles in the Christchurch bus fleet. In particular, bus age was compared with observed bus noisiness, bus arrival time, and bus graffiti and damage. No meaningful correlation between bus noisiness and age was observed, although a slight overall increase in noise levels can be observed with age. There was also no meaningful correlation between bus arrival time and age, although a slight overall increase in lateness can be observed with age. And finally, there was no correlation, dis- cernible relationship, or trend between observed damage and age. Benefit Type Activity Annual User Benefit Reliability improvements Major mechanical overhaul or bus replacement $204,000 Cleanliness Effective cleaning programs $4,230 Facilities Legroom, Sufficient Information Displayed, CCTV, Seating, Air Cond, and other user comforts $10,382 Ride Provide smooth ride $1,692 Total $220,304 Table C-10. User benefits for hypothetical single bus upgrade. Figure C-7. Satisfaction trends vs. 2016 average fleet age in Auckland, Wellington, and Christchurch.

C-16 The Relationship Between Transit Asset Condition and Service Quality References Richter, M.; Ilzig, K.; and Rudnicki, A. (2009). “Models for the Bus Headway Distribution in the Flow Behind a Traffic Signal.” In Proceedings of the 18th International Conference on the Application of Computer Science and Mathematics in Architecture and Civil Engineering, Weimar, Germany. Robert, W.; Reeder, V.; Lawrence, K.; Cohen, H.; and O’Neil, K. (2014) Guidance for Developing the State of Good Repair Prioritization Framework and Tools: Research Report. Contractor’s Final Report for TCRP Project E-09A. TRB. Ruan, Minyan, and Lin, J. (2009). “An Investigation of Bus Headway Regularity and Service Performance in Chicago Bus Transit System.” Transport Chicago, Annual Conference. Wardman, M., and Whelan, G. (2001). Valuation of Improved Railway Rolling Stock: A Review of Literature and New Evidence. Transport Reviews, 21(4), 415–447. https://doi.org/10.1080/01441640010020115